timbal 2.8.2__tar.gz → 2.8.4__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Potentially problematic release.
This version of timbal might be problematic. Click here for more details.
- {timbal-2.8.2 → timbal-2.8.4}/PKG-INFO +1 -1
- {timbal-2.8.2 → timbal-2.8.4}/pyproject.toml +6 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/_version.py +2 -2
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/messages.py +14 -1
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/mcp.py +15 -2
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/memory_compaction.py +1 -1
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/models.py +4 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/llm_base.py +2 -2
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/rubric.py +2 -2
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/injection.py +1 -1
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/judge.py +1 -1
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/topic.py +2 -2
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/rubric.py +1 -1
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/models.yaml +137 -27
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/__init__.py +63 -0
- timbal-2.8.4/python/timbal/tools/_oauth_store.py +267 -0
- timbal-2.8.4/python/timbal/tools/notion.py +1282 -0
- timbal-2.8.4/python/timbal/tools/qlik.py +886 -0
- timbal-2.8.4/python/timbal/tools/snowflake.py +952 -0
- {timbal-2.8.2 → timbal-2.8.4}/.gitignore +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/LICENSE +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/README.md +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/chat-completions.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/create-from-workforce.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/jobs/cancel.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/jobs/create.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/jobs/get.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/jobs/list.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/jobs/retry.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/link.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/list-context-vars.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/list-policies.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/docs/api-reference/ace/unlink.mdx +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/pyrightconfig.json +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/_openai_audio.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/_slots.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/__main__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/cli_utils.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/cst_utils.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/flow.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/format.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/guardrail_specs.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/library.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/model_discovery.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/test.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/tool_discovery.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/add_edge.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/add_guardrail.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/add_mcp.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/add_step.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/add_tool.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/remove_edge.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/remove_guardrail.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/remove_step.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/remove_tool.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/set_config.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/set_param.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/codegen/transformers/set_position.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/harmony_leak.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/anthropic.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/default.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/message.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/openai.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/string.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/impl/timbal.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/collectors/registry.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/agent.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/attachment_limit.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/continuation.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/fallback_model.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/auto.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/chat_completions.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/clients.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/registry.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/responses.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/retry.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/llm/router.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/provider_errors.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/runnable.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/session_scheduler.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/skill.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/test_model.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/tool.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/tool_result_offload.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/tool_set.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/core/workflow.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/errors.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/cli.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/display.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/agent.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/eval_examples.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/eval_flow_validators.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/eval_negations.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/eval_parallel.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/eval_subagent.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/eval_transforms.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/fixtures/test_invalid_validator.yaml +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/models.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/reporters.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/runner.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/utils.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/_spans.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/comparison_base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/contains.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/contains_all.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/contains_any.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/context.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/email.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/ends_with.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/eq.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/gt.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/gte.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/json.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/language.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/length.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/lt.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/lte.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/max_length.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/min_length.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/not_null.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/parallel.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/pattern.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/prompt.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/semantic.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/seq.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/starts_with.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/evals/validators/type.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/apply.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/keywords.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/length.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/moderate.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/pii.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/builtins/secrets.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/judge_llm.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/presets.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/runner.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/testing.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/trace.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/guardrails/types.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/logs.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/_http_session.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/integrations.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/knowledge_bases/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/knowledge_bases/query.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/tool_proxy.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/types.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/platform/utils.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/__main__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/capacity.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/http.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/jobs.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/livekit_call_control.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/livekit_session.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/livekit_sip.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/playground.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/recording_upload.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/rtc.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/single_session.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/telephony.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/voice.html +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/server/voice.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/background.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/config.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/config_loader.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/context.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/dependency_analyzer.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/exporters/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/exporters/otel.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/providers/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/providers/base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/providers/in_memory.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/providers/jsonl.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/providers/platform.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/providers/sqlite.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/span.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/trace.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/state/tracing/uploads.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/_creds.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/aircall.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/asana.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/bash.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/cala.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/cloudflare.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/connectif.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/covermanager.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/deepgram.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/dynamics_business_central.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/dynamics_sales.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/edit.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/elasticsearch.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/elevenlabs.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/excel.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/fal.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/fathom.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/firecrawl.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/freematica.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/freematica_contracts.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/freematica_people.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/freematica_purchasing.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/freematica_sales.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/gemini_images.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/gmail.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_analytics.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_business.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_calendar.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_docs.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_drive.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_maps.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_merchant_center.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_search_console.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/google_sheets.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/happy_scribe.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/higgsfield.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/hubspot.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/interaction.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/jira.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/klaviyo.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/knowledge_base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/krea.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/lancedb.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/linkedin.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/mongodb.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/netsuite.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/onedrive.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/openai_audio.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/outlook.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/parallel.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/pinecone.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/powerbi.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/quiver_ai.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/read.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/replicate.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/resend.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/salesforce.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/scraperapi.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/sharepoint.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/shopify.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/slack.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/stripe.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/tavily.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/web_search.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/write.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/xai.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/zendesk.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/tools/zoho_crm.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/approval.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/custom.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/file.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/text.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/thinking.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/tool_result.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/content/tool_use.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/approval.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/base.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/delta.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/guardrail.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/interaction.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/output.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/events/start.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/file.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/message.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/types/run_status.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/utils/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/utils/import_spec.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/utils/model.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/utils/net.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/utils/schema.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/utils/serialization.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/__init__.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/_whisper_features.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/ambience.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/config.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/deepgram.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/deepgram_tts.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/elevenlabs.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/endpointing.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/eou.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/events.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/fish_audio.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/livekit.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/metrics.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/munsit.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/namo.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/openai.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/playback.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/providers.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/realtime.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/recording.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/session.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/smart_turn.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/telephony.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/turn_detection.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/vad.py +0 -0
- {timbal-2.8.2 → timbal-2.8.4}/python/timbal/voice/webrtc.py +0 -0
|
@@ -129,6 +129,12 @@ path = "python"
|
|
|
129
129
|
[tool.hatch.version]
|
|
130
130
|
source = "vcs"
|
|
131
131
|
|
|
132
|
+
# Versions come from git tags. On a release branch (`release/2.8`) the next version is the next patch
|
|
133
|
+
# (2.8.5.devN after v2.8.4); anywhere else (main, feature branches) it is the next minor (2.9.0.devN),
|
|
134
|
+
# so main never builds with a version a release line will publish. See RELEASING.md.
|
|
135
|
+
[tool.hatch.version.raw-options]
|
|
136
|
+
version_scheme = "release-branch-semver"
|
|
137
|
+
|
|
132
138
|
[tool.hatch.build.hooks.vcs]
|
|
133
139
|
version-file = "python/timbal/_version.py"
|
|
134
140
|
|
|
@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
|
|
|
18
18
|
commit_id: str | None
|
|
19
19
|
__commit_id__: str | None
|
|
20
20
|
|
|
21
|
-
__version__ = version = '2.8.
|
|
22
|
-
__version_tuple__ = version_tuple = (2, 8,
|
|
21
|
+
__version__ = version = '2.8.4'
|
|
22
|
+
__version_tuple__ = version_tuple = (2, 8, 4)
|
|
23
23
|
|
|
24
24
|
__commit_id__ = commit_id = None
|
|
@@ -1,5 +1,6 @@
|
|
|
1
1
|
"""Anthropic Messages API request adapter (anthropic)."""
|
|
2
2
|
|
|
3
|
+
import re
|
|
3
4
|
from collections.abc import AsyncIterator, Callable
|
|
4
5
|
from typing import TYPE_CHECKING, Any
|
|
5
6
|
|
|
@@ -12,6 +13,16 @@ if TYPE_CHECKING:
|
|
|
12
13
|
from ..runnable import Runnable
|
|
13
14
|
|
|
14
15
|
|
|
16
|
+
# Claude models whose API rejects non-default sampling parameters with a 400 (Opus 4.7+, the Opus/Sonnet/Haiku 5
|
|
17
|
+
# generations, Fable and Mythos). Anthropic's model docs say to omit `temperature`, `top_p` and `top_k` for them.
|
|
18
|
+
_NO_SAMPLING_PARAMS = re.compile(r"^claude-(?:(?:fable|mythos)-|opus-(?:4-[78]|5)|sonnet-5|haiku-5)")
|
|
19
|
+
|
|
20
|
+
|
|
21
|
+
def rejects_temperature(model_name: str) -> bool:
|
|
22
|
+
"""Whether the Anthropic API rejects a ``temperature`` value for this model (dated ids included)."""
|
|
23
|
+
return _NO_SAMPLING_PARAMS.match(model_name) is not None
|
|
24
|
+
|
|
25
|
+
|
|
15
26
|
def prepare_messages_request(
|
|
16
27
|
*,
|
|
17
28
|
client: Any,
|
|
@@ -41,7 +52,9 @@ def prepare_messages_request(
|
|
|
41
52
|
if anthropic_tools:
|
|
42
53
|
anthropic_kwargs["tools"] = anthropic_tools
|
|
43
54
|
|
|
44
|
-
|
|
55
|
+
# Deterministic callers (LLM judges pin temperature=0.0) would 400 on these models, so the value is dropped
|
|
56
|
+
# for them. Sending it anyway via ``model_params={"temperature": ...}`` still reaches the API unchanged.
|
|
57
|
+
if temperature is not None and not rejects_temperature(model_name):
|
|
45
58
|
anthropic_kwargs["temperature"] = temperature
|
|
46
59
|
|
|
47
60
|
# Server-side automatic prompt caching (anthropic SDK >= 0.83): the API
|
|
@@ -363,6 +363,15 @@ class MCPServer(ToolSet):
|
|
|
363
363
|
url: str | None = None
|
|
364
364
|
headers: dict[str, str] = Field(default_factory=dict)
|
|
365
365
|
|
|
366
|
+
auth: httpx.Auth | None = Field(default=None, exclude=True)
|
|
367
|
+
"""``httpx.Auth`` applied to every HTTP request, for credentials that rotate during a
|
|
368
|
+
session (``headers`` are fixed at connect). E.g. ``SnowflakeAuth`` re-mints its
|
|
369
|
+
key-pair JWT before the one-hour expiry and retries once on 401.
|
|
370
|
+
|
|
371
|
+
The auth flow runs in the transport's owner task, which has an empty context (no
|
|
372
|
+
``RunContext``). If the object defines ``async prepare()``, it is awaited in the
|
|
373
|
+
caller's context before every request; resolve anything identity-bound there."""
|
|
374
|
+
|
|
366
375
|
timeout: float | None = DEFAULT_TIMEOUT_SECONDS
|
|
367
376
|
"""Seconds to wait for the server to answer a request on an established session
|
|
368
377
|
(``tools/list``, ``tools/call``). Defaults to ``DEFAULT_TIMEOUT_SECONDS`` (10 minutes);
|
|
@@ -439,8 +448,8 @@ class MCPServer(ToolSet):
|
|
|
439
448
|
@model_validator(mode="after")
|
|
440
449
|
def _validate_transport_fields(self) -> "MCPServer":
|
|
441
450
|
if self.service is not None:
|
|
442
|
-
if self.transport != "http" or self.url is not None or self.headers:
|
|
443
|
-
raise ValueError("service='api' requires HTTP transport without url or
|
|
451
|
+
if self.transport != "http" or self.url is not None or self.headers or self.auth is not None:
|
|
452
|
+
raise ValueError("service='api' requires HTTP transport without url, headers or auth")
|
|
444
453
|
if self.transport == "stdio":
|
|
445
454
|
if not self.command:
|
|
446
455
|
raise ValueError("'command' is required for stdio transport")
|
|
@@ -520,6 +529,7 @@ class MCPServer(ToolSet):
|
|
|
520
529
|
assert self.url is not None
|
|
521
530
|
async with httpx.AsyncClient(
|
|
522
531
|
headers=self.headers if self.headers else None,
|
|
532
|
+
auth=self.auth,
|
|
523
533
|
timeout=self._http_timeout(),
|
|
524
534
|
follow_redirects=True,
|
|
525
535
|
) as http_client:
|
|
@@ -720,6 +730,9 @@ class MCPServer(ToolSet):
|
|
|
720
730
|
"""
|
|
721
731
|
if timeout is _USE_SERVER_TIMEOUT:
|
|
722
732
|
timeout = self.timeout
|
|
733
|
+
prepare = getattr(self.auth, "prepare", None)
|
|
734
|
+
if prepare is not None:
|
|
735
|
+
await prepare()
|
|
723
736
|
for attempt in (1, 2):
|
|
724
737
|
session = await self._connect()
|
|
725
738
|
owner = self._session_task
|
|
@@ -715,7 +715,7 @@ def summarize(
|
|
|
715
715
|
threshold: Trigger summarization when message count exceeds this.
|
|
716
716
|
model: Model for summarization. Defaults to None, which uses the
|
|
717
717
|
agent's own model. Override with a cheaper model like
|
|
718
|
-
'
|
|
718
|
+
'anthropic/claude-haiku-5-5' to reduce cost.
|
|
719
719
|
keep_last_n: Number of recent messages to keep unsummarized.
|
|
720
720
|
max_summary_tokens: Maximum tokens for the summary response.
|
|
721
721
|
preserve_user_messages: Carry the user's messages from the summarized region
|
|
@@ -148,6 +148,7 @@ def has_cache_write_pricing(model_id: str) -> bool:
|
|
|
148
148
|
# ---------------------------------------------------------------------------
|
|
149
149
|
# Model type with provider prefixes
|
|
150
150
|
Model = Literal[
|
|
151
|
+
"anthropic/claude-fable-5-2",
|
|
151
152
|
"anthropic/claude-fable-5-1",
|
|
152
153
|
"anthropic/claude-fable-5",
|
|
153
154
|
"anthropic/claude-opus-5-5",
|
|
@@ -162,8 +163,11 @@ Model = Literal[
|
|
|
162
163
|
"anthropic/claude-sonnet-4-5",
|
|
163
164
|
"anthropic/claude-haiku-5-5",
|
|
164
165
|
"anthropic/claude-haiku-4-5",
|
|
166
|
+
"openai/gpt-6.2-astra",
|
|
167
|
+
"openai/gpt-6.2-sol",
|
|
165
168
|
"openai/gpt-6-astra",
|
|
166
169
|
"openai/gpt-6.1-sol",
|
|
170
|
+
"openai/gpt-6.1-terra",
|
|
167
171
|
"openai/gpt-6-sol",
|
|
168
172
|
"openai/gpt-6-luna",
|
|
169
173
|
"openai/gpt-5.5",
|
|
@@ -22,7 +22,7 @@ class LLMValidator(BaseValidator):
|
|
|
22
22
|
Attributes:
|
|
23
23
|
system_prompt: System prompt for the LLM. Can be customized in YAML.
|
|
24
24
|
Subclasses should provide a default value.
|
|
25
|
-
model: LLM model to use for validation. Defaults to "
|
|
25
|
+
model: LLM model to use for validation. Defaults to "anthropic/claude-haiku-5-5".
|
|
26
26
|
Can be customized in YAML.
|
|
27
27
|
|
|
28
28
|
Subclasses should implement:
|
|
@@ -31,7 +31,7 @@ class LLMValidator(BaseValidator):
|
|
|
31
31
|
"""
|
|
32
32
|
|
|
33
33
|
system_prompt: str # Must be set in subclass or YAML
|
|
34
|
-
model: str = "
|
|
34
|
+
model: str = "anthropic/claude-haiku-5-5" # Can be overridden in YAML
|
|
35
35
|
|
|
36
36
|
@abstractmethod
|
|
37
37
|
def get_user_prompt(self, actual_value: Any) -> str:
|
|
@@ -33,7 +33,7 @@ class RubricValidator(BaseValidator):
|
|
|
33
33
|
name: recommendations
|
|
34
34
|
weight: 2
|
|
35
35
|
pass_threshold: 0.75
|
|
36
|
-
model:
|
|
36
|
+
model: anthropic/claude-haiku-5-5
|
|
37
37
|
context: "The agent produced a price-comparison report."
|
|
38
38
|
|
|
39
39
|
# Markdown rubric (bullet lines become criteria):
|
|
@@ -47,7 +47,7 @@ class RubricValidator(BaseValidator):
|
|
|
47
47
|
"""
|
|
48
48
|
|
|
49
49
|
name: Literal["rubric!"] = "rubric!" # type: ignore
|
|
50
|
-
model: SkipValidation[Any] = "
|
|
50
|
+
model: SkipValidation[Any] = "anthropic/claude-haiku-5-5"
|
|
51
51
|
"""Judge model. Use a small, cheap model."""
|
|
52
52
|
pass_threshold: float = 1.0
|
|
53
53
|
"""Weighted fraction of criteria that must pass (1.0 = all)."""
|
|
@@ -64,7 +64,7 @@ class PromptInjection(Guardrail):
|
|
|
64
64
|
stages: set[GuardrailStage] = Field(default_factory=lambda: {GuardrailStage.INPUT})
|
|
65
65
|
action: str = "block"
|
|
66
66
|
model: Any = None
|
|
67
|
-
"""Optional LLM classifier model — a model string (e.g. '
|
|
67
|
+
"""Optional LLM classifier model — a model string (e.g. 'anthropic/claude-haiku-5-5'), or any
|
|
68
68
|
model instance the router accepts (FallbackModel, TestModel). None = patterns only."""
|
|
69
69
|
max_classifier_chars: int = 8_000
|
|
70
70
|
"""Input beyond this is truncated before classification (patterns still see all of it)."""
|
|
@@ -58,7 +58,7 @@ class LLMJudge(Guardrail):
|
|
|
58
58
|
"""Rubric mode: weighted fraction of criteria that must pass (1.0 = all)."""
|
|
59
59
|
rubric_context: str | None = None
|
|
60
60
|
"""Optional task description shown to every rubric judge."""
|
|
61
|
-
model: Any = "
|
|
61
|
+
model: Any = "anthropic/claude-haiku-5-5"
|
|
62
62
|
max_chars: int = 16_000
|
|
63
63
|
|
|
64
64
|
def __init__(self, criteria: str | None = None, **kwargs: Any) -> None:
|
|
@@ -21,7 +21,7 @@ class TopicGuard(Guardrail):
|
|
|
21
21
|
```python
|
|
22
22
|
TopicGuard(allow=["billing", "shipping"],
|
|
23
23
|
blocked_message="I can only help with billing and shipping.")
|
|
24
|
-
TopicGuard(deny=["medical advice", "legal advice"], model="
|
|
24
|
+
TopicGuard(deny=["medical advice", "legal advice"], model="anthropic/claude-haiku-5-5")
|
|
25
25
|
```
|
|
26
26
|
"""
|
|
27
27
|
|
|
@@ -32,7 +32,7 @@ class TopicGuard(Guardrail):
|
|
|
32
32
|
"""Topics the agent may discuss. Anything else is off-topic."""
|
|
33
33
|
deny: list[str] = Field(default_factory=list)
|
|
34
34
|
"""Topics that are always off-topic (checked in addition to allow)."""
|
|
35
|
-
model: Any = "
|
|
35
|
+
model: Any = "anthropic/claude-haiku-5-5"
|
|
36
36
|
"""Classifier model: a model string, or any model instance the router accepts."""
|
|
37
37
|
max_chars: int = 8_000
|
|
38
38
|
|
|
@@ -31,11 +31,31 @@
|
|
|
31
31
|
# ultrafast: 6.0
|
|
32
32
|
|
|
33
33
|
models:
|
|
34
|
+
- id: anthropic/claude-fable-5-2
|
|
35
|
+
provider: anthropic
|
|
36
|
+
display_name: Claude Fable 5.2
|
|
37
|
+
description: Anthropic's most capable generally available model — successor to Fable 5.1 for long-running agentic coding,
|
|
38
|
+
knowledge work, and research, at a 20% lower price.
|
|
39
|
+
notes: 'Released Nov 9, 2026; verified 2026-11-20 against the model reference and pricing. Adaptive thinking is always on
|
|
40
|
+
(disabled/budgeted thinking return 400); effort accepts low|medium|high|xhigh|max, default high. `tool_choice`
|
|
41
|
+
any/tool return 400 — use strict tools or structured outputs instead. Omit temperature, top_p, top_k. Cache reads
|
|
42
|
+
are $0.20 per 1M tokens (0.025x input). Batch is $4/$20. Fast mode (Claude API) is $16/$80 and cannot be combined
|
|
43
|
+
with Batch. 1M context at standard rates, 128K max output, Jun 2026 knowledge cutoff. Requires 30-day data retention;
|
|
44
|
+
not available under ZDR. Refusals return stop_reason refusal; server-side fallback targets are Opus 4.8 and Opus 5.'
|
|
45
|
+
input_price: 8.0
|
|
46
|
+
output_price: 40.0
|
|
47
|
+
cached_input_price: 0.2
|
|
48
|
+
cache_write_price: 10.0
|
|
49
|
+
cache_write_1h_price: 16.0
|
|
50
|
+
service_tiers:
|
|
51
|
+
fast: 2.0
|
|
52
|
+
context_window: 1000000
|
|
53
|
+
capabilities: [vision, tools, reasoning]
|
|
34
54
|
- id: anthropic/claude-fable-5-1
|
|
35
55
|
provider: anthropic
|
|
36
56
|
display_name: Claude Fable 5.1
|
|
37
|
-
description:
|
|
38
|
-
|
|
57
|
+
description: Previous Fable generation for long-running agentic coding, knowledge work, and research, with fewer
|
|
58
|
+
false-positive safety refusals in multi-turn agent loops. Superseded by Fable 5.2 at a 20% lower price.
|
|
39
59
|
notes: Adaptive thinking is always on. Cache reads are $0.25 per 1M tokens (0.025x input).
|
|
40
60
|
`tool_choice` any/tool return 400 — use strict tools or structured outputs instead. Requires 30-day data retention;
|
|
41
61
|
not available under ZDR.
|
|
@@ -167,6 +187,7 @@ models:
|
|
|
167
187
|
provider: anthropic
|
|
168
188
|
display_name: Claude Opus 4.5
|
|
169
189
|
description: Previous-generation Opus model with high intelligence for complex reasoning tasks.
|
|
190
|
+
notes: "Status checked 2026-11-20 against Anthropic's deprecations page: active, retirement not sooner than November 24, 2026. Prefer claude-opus-5-5 or claude-opus-4-8."
|
|
170
191
|
input_price: 5.0
|
|
171
192
|
output_price: 25.0
|
|
172
193
|
cached_input_price: 0.5
|
|
@@ -227,6 +248,7 @@ models:
|
|
|
227
248
|
display_name: Claude Haiku 4.5
|
|
228
249
|
description: Previous Haiku generation for office document processing and high-volume business analysis. Superseded by
|
|
229
250
|
Haiku 5.5.
|
|
251
|
+
notes: "Status checked 2026-11-20 against Anthropic's deprecations page: listed active with retirement 'not sooner than October 15, 2026', a date that has already passed, so retirement may be announced at any time. Prefer claude-haiku-5-5."
|
|
230
252
|
input_price: 1.0
|
|
231
253
|
output_price: 5.0
|
|
232
254
|
cached_input_price: 0.1
|
|
@@ -234,11 +256,66 @@ models:
|
|
|
234
256
|
cache_write_1h_price: 2.0
|
|
235
257
|
context_window: 200000
|
|
236
258
|
capabilities: [vision, tools, reasoning]
|
|
259
|
+
- id: openai/gpt-6.2-astra
|
|
260
|
+
provider: openai
|
|
261
|
+
display_name: GPT-6.2 Astra
|
|
262
|
+
description: OpenAI's flagship model for the hardest end-to-end work — complex reasoning, agentic coding, computer
|
|
263
|
+
use, research, and document creation. Successor to GPT-6 Astra at 20% lower prices.
|
|
264
|
+
notes: 'Released Nov 17, 2026; verified 2026-11-20 against the model reference and API pricing. Standard short-context
|
|
265
|
+
pricing is $8/$40 per 1M tokens; cached input $0.40, 30m cache writes $10 (12h writes $20). Prompts over 272K input
|
|
266
|
+
tokens bill the full request at long-context rates ($16/$60, cached $0.80, cache writes $20). Batch is 50% of
|
|
267
|
+
Standard; Fast mode (`service_tier: "fast"`) is 2x. Ultrafast is promotionally 3x through Feb 28, 2027 and 6x from
|
|
268
|
+
Mar 1, 2027 (the 6x list rate is catalogued). Flex is not supported. `reasoning.effort` accepts
|
|
269
|
+
low|medium|high|xhigh|max (`none` returns 400; no default published — start with medium). Function tools require
|
|
270
|
+
the Responses API (Timbal default). 1.05M context (922K max input), 128K max output, May 18, 2026 knowledge cutoff.
|
|
271
|
+
Supports ZDR for approved organizations. Regional processing adds 10%.'
|
|
272
|
+
input_price: 8.0
|
|
273
|
+
output_price: 40.0
|
|
274
|
+
cached_input_price: 0.4
|
|
275
|
+
cache_write_price: 10.0
|
|
276
|
+
long_context:
|
|
277
|
+
threshold: 272000
|
|
278
|
+
input_price: 16.0
|
|
279
|
+
output_price: 60.0
|
|
280
|
+
cached_input_price: 0.8
|
|
281
|
+
cache_write_price: 20.0
|
|
282
|
+
service_tiers:
|
|
283
|
+
fast: 2.0
|
|
284
|
+
ultrafast: 6.0
|
|
285
|
+
context_window: 1050000
|
|
286
|
+
capabilities: [vision, tools, reasoning]
|
|
287
|
+
- id: openai/gpt-6.2-sol
|
|
288
|
+
provider: openai
|
|
289
|
+
display_name: GPT-6.2 Sol
|
|
290
|
+
description: OpenAI model for complex coding, computer use, and professional work at a lower cost than Astra.
|
|
291
|
+
notes: 'Released Nov 17, 2026; verified 2026-11-20 against the model reference and API pricing. Standard rates per 1M
|
|
292
|
+
tokens are $2 input, $0.10 cache reads, $2.50 cache writes (30m; 12h writes $5), and $10 output. Above 272K input
|
|
293
|
+
tokens the full request costs $4/$15, cached $0.20, cache writes $5. Batch is 50% of Standard; Fast costs 2x.
|
|
294
|
+
Ultrafast is promotionally 3x through Feb 28, 2027 and 6x from Mar 1, 2027 (the 6x list rate is catalogued).
|
|
295
|
+
Flex is not supported. Reasoning accepts low, medium, high, xhigh, max; `none` returns 400 and no default is
|
|
296
|
+
published (start with medium). Tool calling requires Responses; Chat Completions supports text without tools.
|
|
297
|
+
1.05M context (922K max input), 128K max output, Jun 1, 2026 knowledge cutoff. Supports ZDR for approved
|
|
298
|
+
organizations. Regional processing adds 10%.'
|
|
299
|
+
input_price: 2.0
|
|
300
|
+
output_price: 10.0
|
|
301
|
+
cached_input_price: 0.1
|
|
302
|
+
cache_write_price: 2.5
|
|
303
|
+
long_context:
|
|
304
|
+
threshold: 272000
|
|
305
|
+
input_price: 4.0
|
|
306
|
+
output_price: 15.0
|
|
307
|
+
cached_input_price: 0.2
|
|
308
|
+
cache_write_price: 5.0
|
|
309
|
+
service_tiers:
|
|
310
|
+
fast: 2.0
|
|
311
|
+
ultrafast: 6.0
|
|
312
|
+
context_window: 1050000
|
|
313
|
+
capabilities: [vision, tools, reasoning]
|
|
237
314
|
- id: openai/gpt-6-astra
|
|
238
315
|
provider: openai
|
|
239
316
|
display_name: GPT-6 Astra
|
|
240
|
-
description:
|
|
241
|
-
use, research, and document creation. Successor to GPT-5.6 Sol.
|
|
317
|
+
description: Previous Astra generation for the hardest end-to-end work — complex reasoning, agentic coding, computer
|
|
318
|
+
use, research, and document creation. Successor to GPT-5.6 Sol; superseded by GPT-6.2 Astra at 20% lower prices.
|
|
242
319
|
notes: 'Standard short-context pricing is $10/$50 per 1M tokens; cached input $1 (0.1x), cache writes $12.50 (1.25x). Prompts
|
|
243
320
|
over 272K input tokens bill the full request at long-context rates ($20/$75, cached $2, cache writes $25). Batch/Flex are
|
|
244
321
|
50% of Standard; Fast mode (`service_tier: "fast"`) is 2x; Ultrafast is 6x. `reasoning.effort` accepts low|medium|high|xhigh|max (API
|
|
@@ -264,7 +341,7 @@ models:
|
|
|
264
341
|
- id: openai/gpt-6.1-sol
|
|
265
342
|
provider: openai
|
|
266
343
|
display_name: GPT-6.1 Sol
|
|
267
|
-
description:
|
|
344
|
+
description: Previous Sol generation for complex coding, computer use, and professional work. Superseded by GPT-6.2 Sol at the same price.
|
|
268
345
|
notes: 'Verified 2026-10-03 against the model card and API pricing. Standard rates per 1M tokens are $2 input,
|
|
269
346
|
$0.10 cache reads, $2.50 cache writes, and $10 output. Above 272K input tokens the full request costs
|
|
270
347
|
$4/$15, cached $0.20, cache writes $5. Batch/Flex cost 50% of Standard; Fast costs 2x.
|
|
@@ -289,6 +366,32 @@ models:
|
|
|
289
366
|
ultrafast: 6.0
|
|
290
367
|
context_window: 1050000
|
|
291
368
|
capabilities: [vision, tools, reasoning]
|
|
369
|
+
- id: openai/gpt-6.1-terra
|
|
370
|
+
provider: openai
|
|
371
|
+
display_name: GPT-6.1 Terra
|
|
372
|
+
description: OpenAI model for fast, efficient visual reasoning, precise computer use, and clear writing, priced between Sol and Luna.
|
|
373
|
+
notes: 'Verified 2026-11-20 against the model reference and API pricing (release date not published there). Standard rates
|
|
374
|
+
per 1M tokens are $1 input, $0.05 cache reads, $1.25 cache writes (30m; 12h writes $2.50), and $5 output. Above 272K
|
|
375
|
+
input tokens the full request costs $2/$7.50, cached $0.10, cache writes $2.50. Batch/Flex cost 50% of Standard (Flex is
|
|
376
|
+
being wound down and will not be added to new models); Fast costs 2x; Ultrafast is not offered. Reasoning accepts
|
|
377
|
+
none, low, medium (default), high, xhigh, max. Tool calling requires Responses; Chat Completions supports text without
|
|
378
|
+
tools. 1.05M context (922K max input), 128K max output, Jun 1, 2026 knowledge cutoff. Regional processing adds 10%
|
|
379
|
+
where eligible. ZDR support is not stated on the model page.'
|
|
380
|
+
input_price: 1.0
|
|
381
|
+
output_price: 5.0
|
|
382
|
+
cached_input_price: 0.05
|
|
383
|
+
cache_write_price: 1.25
|
|
384
|
+
long_context:
|
|
385
|
+
threshold: 272000
|
|
386
|
+
input_price: 2.0
|
|
387
|
+
output_price: 7.5
|
|
388
|
+
cached_input_price: 0.1
|
|
389
|
+
cache_write_price: 2.5
|
|
390
|
+
service_tiers:
|
|
391
|
+
flex: 0.5
|
|
392
|
+
fast: 2.0
|
|
393
|
+
context_window: 1050000
|
|
394
|
+
capabilities: [vision, tools, reasoning]
|
|
292
395
|
- id: openai/gpt-6-sol
|
|
293
396
|
provider: openai
|
|
294
397
|
display_name: GPT-6 Sol
|
|
@@ -370,8 +473,8 @@ models:
|
|
|
370
473
|
provider: openai
|
|
371
474
|
display_name: GPT-5.6 Sol
|
|
372
475
|
description: OpenAI GPT-5.6 flagship tier for the hardest coding, agentic, and long-horizon tasks.
|
|
373
|
-
notes: 'Promotional rates retained because post-promotion rates are not published.
|
|
374
|
-
|
|
476
|
+
notes: 'Promotional rates retained because post-promotion rates are not published. Promotional pricing
|
|
477
|
+
is available until November 30, 2026 (per the OpenAI pricing page; re-verify after that date): $4/$20 input/output per 1M tokens, cached input $0.40, cache writes $5.
|
|
375
478
|
Above 272K input tokens the full request costs $8/$30, cached input $0.80, cache writes $10.'
|
|
376
479
|
input_price: 4.0
|
|
377
480
|
output_price: 20.0
|
|
@@ -498,6 +601,7 @@ models:
|
|
|
498
601
|
provider: openai
|
|
499
602
|
display_name: GPT-5
|
|
500
603
|
description: OpenAI GPT-5 base model.
|
|
604
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Deprecated (notice June 11, 2026); shuts down December 11, 2026, including its dated snapshot. Prefer gpt-6.1-sol."
|
|
501
605
|
input_price: 1.25
|
|
502
606
|
output_price: 10.0
|
|
503
607
|
cached_input_price: 0.125
|
|
@@ -507,6 +611,7 @@ models:
|
|
|
507
611
|
provider: openai
|
|
508
612
|
display_name: GPT-5 Mini
|
|
509
613
|
description: Compact GPT-5 variant for everyday tasks.
|
|
614
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Deprecated (notice June 11, 2026); shuts down December 11, 2026, including its dated snapshot. Prefer gpt-6.1-terra."
|
|
510
615
|
input_price: 0.25
|
|
511
616
|
output_price: 2.0
|
|
512
617
|
cached_input_price: 0.025
|
|
@@ -516,6 +621,7 @@ models:
|
|
|
516
621
|
provider: openai
|
|
517
622
|
display_name: GPT-5 Nano
|
|
518
623
|
description: Ultra-lightweight GPT-5 variant for cost-sensitive use cases.
|
|
624
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Deprecated (notice June 11, 2026); shuts down December 11, 2026, including its dated snapshot. Prefer gpt-6-luna."
|
|
519
625
|
input_price: 0.05
|
|
520
626
|
output_price: 0.4
|
|
521
627
|
cached_input_price: 0.005
|
|
@@ -525,6 +631,7 @@ models:
|
|
|
525
631
|
provider: openai
|
|
526
632
|
display_name: GPT-4.1
|
|
527
633
|
description: Improved instruction-following and multimodal model with a 1M token context window.
|
|
634
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Dated snapshot deprecated (notice October 14, 2026); shuts down January 14, 2027. The bare alias is not named on the page. Prefer gpt-6-luna."
|
|
528
635
|
input_price: 2.0
|
|
529
636
|
output_price: 8.0
|
|
530
637
|
cached_input_price: 0.5
|
|
@@ -534,6 +641,7 @@ models:
|
|
|
534
641
|
provider: openai
|
|
535
642
|
display_name: GPT-4.1 Mini
|
|
536
643
|
description: Compact GPT-4.1 variant for everyday multimodal tasks.
|
|
644
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Dated snapshot deprecated (notice October 14, 2026); shuts down January 14, 2027. The bare alias is not named on the page. Prefer gpt-6-luna."
|
|
537
645
|
input_price: 0.4
|
|
538
646
|
output_price: 1.6
|
|
539
647
|
cached_input_price: 0.1
|
|
@@ -543,7 +651,7 @@ models:
|
|
|
543
651
|
provider: openai
|
|
544
652
|
display_name: GPT-4.1 Nano
|
|
545
653
|
description: Smallest GPT-4.1 variant for high-volume, cost-sensitive workloads.
|
|
546
|
-
notes:
|
|
654
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Shut down October 23, 2026 (replacement: gpt-5.6-luna); requests are expected to fail. Kept only so existing references still type-check."
|
|
547
655
|
input_price: 0.1
|
|
548
656
|
output_price: 0.4
|
|
549
657
|
cached_input_price: 0.025
|
|
@@ -553,6 +661,7 @@ models:
|
|
|
553
661
|
provider: openai
|
|
554
662
|
display_name: GPT-4o
|
|
555
663
|
description: Omni model combining vision, audio, and text understanding in a 128K context.
|
|
664
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Snapshots gpt-4o-2024-08-06 and gpt-4o-2024-11-20 are deprecated and shut down January 14, 2027 (gpt-4o-2024-05-13 shut down October 23, 2026). The bare alias is not named on the page. Prefer gpt-6-luna."
|
|
556
665
|
input_price: 2.5
|
|
557
666
|
output_price: 10.0
|
|
558
667
|
cached_input_price: 1.25
|
|
@@ -562,6 +671,7 @@ models:
|
|
|
562
671
|
provider: openai
|
|
563
672
|
display_name: GPT-4o Mini
|
|
564
673
|
description: Fast, affordable small model for focused tasks.
|
|
674
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Snapshot gpt-4o-mini-2024-07-18 is deprecated and shuts down January 14, 2027. The bare alias is not named on the page. Prefer gpt-6-luna."
|
|
565
675
|
input_price: 0.15
|
|
566
676
|
output_price: 0.6
|
|
567
677
|
cached_input_price: 0.075
|
|
@@ -571,7 +681,7 @@ models:
|
|
|
571
681
|
provider: openai
|
|
572
682
|
display_name: o4 Mini
|
|
573
683
|
description: Compact reasoning model balancing speed and deep thinking.
|
|
574
|
-
notes:
|
|
684
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Shut down October 23, 2026 (replacement: gpt-5.6-terra); requests are expected to fail. Kept only so existing references still type-check."
|
|
575
685
|
input_price: 1.1
|
|
576
686
|
output_price: 4.4
|
|
577
687
|
cached_input_price: 0.275
|
|
@@ -581,7 +691,7 @@ models:
|
|
|
581
691
|
provider: openai
|
|
582
692
|
display_name: o3
|
|
583
693
|
description: Advanced reasoning model for complex math, science, and coding.
|
|
584
|
-
notes: Deprecated;
|
|
694
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Deprecated (notice June 11, 2026); o3 and o3-2025-04-16 shut down December 11, 2026. Prefer gpt-6.1-sol."
|
|
585
695
|
input_price: 2.0
|
|
586
696
|
output_price: 8.0
|
|
587
697
|
cached_input_price: 0.5
|
|
@@ -591,7 +701,7 @@ models:
|
|
|
591
701
|
provider: openai
|
|
592
702
|
display_name: o3 Mini
|
|
593
703
|
description: Compact o3 variant for efficient deep reasoning.
|
|
594
|
-
notes:
|
|
704
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Shut down October 23, 2026 (replacement: gpt-5.6-sol); requests are expected to fail. Kept only so existing references still type-check."
|
|
595
705
|
input_price: 1.1
|
|
596
706
|
output_price: 4.4
|
|
597
707
|
cached_input_price: 0.55
|
|
@@ -601,7 +711,7 @@ models:
|
|
|
601
711
|
provider: openai
|
|
602
712
|
display_name: o3 Pro
|
|
603
713
|
description: High-compute o3 variant for the most demanding reasoning tasks.
|
|
604
|
-
notes: Deprecated;
|
|
714
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Deprecated (notice June 11, 2026); o3-pro and o3-pro-2025-06-10 shut down December 11, 2026. Prefer gpt-6.1-sol with reasoning.mode=pro (Responses API)."
|
|
605
715
|
input_price: 20.0
|
|
606
716
|
output_price: 80.0
|
|
607
717
|
context_window: 200000
|
|
@@ -610,7 +720,7 @@ models:
|
|
|
610
720
|
provider: openai
|
|
611
721
|
display_name: o1
|
|
612
722
|
description: First-generation reasoning model with extended thinking for complex problems.
|
|
613
|
-
notes:
|
|
723
|
+
notes: "Status checked 2026-11-20 against OpenAI's deprecations page. Shut down October 23, 2026 (replacement: gpt-5.6-sol); requests are expected to fail. Kept only so existing references still type-check."
|
|
614
724
|
input_price: 15.0
|
|
615
725
|
output_price: 60.0
|
|
616
726
|
cached_input_price: 7.5
|
|
@@ -635,6 +745,8 @@ models:
|
|
|
635
745
|
provider: togetherai
|
|
636
746
|
display_name: Llama 3.3 70B Instruct Turbo
|
|
637
747
|
description: Multilingual 70B model delivering enhanced performance and matching Llama 3.2 90B on text-only tasks.
|
|
748
|
+
notes: "Together's deprecations page (checked 2026-11-20) lists this model as removed from serverless on 2026-10-22; dedicated endpoint only. Probe: 'non-serverless model'."
|
|
749
|
+
dedicated_only: true
|
|
638
750
|
input_price: 1.04
|
|
639
751
|
output_price: 1.04
|
|
640
752
|
context_window: 131072
|
|
@@ -767,6 +879,7 @@ models:
|
|
|
767
879
|
provider: togetherai
|
|
768
880
|
display_name: MiniMax M3 (Together)
|
|
769
881
|
description: MiniMax M3 multimodal model on Together for chat, agents, and long-context workloads.
|
|
882
|
+
notes: "Together's deprecations page (checked 2026-11-20) lists this model as removed from serverless on 2026-11-06 (replacement MiniMaxAI/MiniMax-M3.1-Flash, dedicated). Probe: 'endpoint has no deployments configured'."
|
|
770
883
|
input_price: 0.3
|
|
771
884
|
output_price: 1.2
|
|
772
885
|
context_window: 524288
|
|
@@ -834,6 +947,7 @@ models:
|
|
|
834
947
|
provider: togetherai
|
|
835
948
|
display_name: DeepSeek V4 Flash 0731 (Together)
|
|
836
949
|
description: DeepSeek V4 Flash 0731 on Together serverless.
|
|
950
|
+
notes: "Together's deprecations page (checked 2026-11-20) lists this model as removed from serverless on 2026-10-22. Probe: 'endpoint has no deployments configured'."
|
|
837
951
|
input_price: 0.14
|
|
838
952
|
output_price: 0.28
|
|
839
953
|
context_window: 1048576
|
|
@@ -902,8 +1016,7 @@ models:
|
|
|
902
1016
|
output_price: 6.0
|
|
903
1017
|
context_window: 1000000
|
|
904
1018
|
capabilities: [vision, tools, reasoning]
|
|
905
|
-
notes: 'Cached-input pricing unresolved: serverless docs quote $0.50/1M, while pricing and model pages quote $0.25/1M. No
|
|
906
|
-
cached rate is assigned until confirmed. https://docs.together.ai/docs/serverless/models and https://www.together.ai/models/qwen3-8-max'
|
|
1019
|
+
notes: "Together's deprecations page (checked 2026-11-20) lists this model as removed from serverless on 2026-11-04; not available dedicated either. Probe: 'non-serverless model'. Cached-input pricing unresolved: serverless docs quote $0.50/1M, while pricing and model pages quote $0.25/1M. No cached rate is assigned until confirmed. https://docs.together.ai/docs/serverless/models and https://www.together.ai/models/qwen3-8-max"
|
|
907
1020
|
- id: google/gemini-4-argon
|
|
908
1021
|
provider: google
|
|
909
1022
|
display_name: Gemini 4 Argon
|
|
@@ -1040,8 +1153,7 @@ models:
|
|
|
1040
1153
|
provider: google
|
|
1041
1154
|
display_name: Gemini 2.5 Pro
|
|
1042
1155
|
description: Legacy Pro model for reasoning, coding, and multimodal analysis; available to previous users.
|
|
1043
|
-
notes:
|
|
1044
|
-
request uses the long-context tier. Cache storage is separate.'
|
|
1156
|
+
notes: "Shuts down December 1, 2026; requests then route to gemini-3.8-flash (Gemini deprecations page, checked 2026-11-20). Not selectable in AI Studio; API access restricted to existing users since 2026-09-18. Above 200,000 input tokens the full request uses the long-context tier. Cache storage is separate."
|
|
1045
1157
|
input_price: 1.25
|
|
1046
1158
|
output_price: 10.0
|
|
1047
1159
|
cached_input_price: 0.125
|
|
@@ -1057,7 +1169,7 @@ models:
|
|
|
1057
1169
|
display_name: Gemini 2.5 Pro TTS
|
|
1058
1170
|
description: Text-to-speech variant of Gemini 2.5 Pro for generating audio output from text input.
|
|
1059
1171
|
notes: 'Native Gemini TTS API required; not supported by the Agent chat route. Output is audio at $20 per 1M audio
|
|
1060
|
-
tokens, not a text output rate. 16,384 max output tokens. Shuts down 2026-
|
|
1172
|
+
tokens, not a text output rate. 16,384 max output tokens. Shuts down 2026-12-18; migrate to
|
|
1061
1173
|
gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts. https://ai.google.dev/gemini-api/docs/deprecations'
|
|
1062
1174
|
input_price: 1.0
|
|
1063
1175
|
output_price: null
|
|
@@ -1067,8 +1179,7 @@ models:
|
|
|
1067
1179
|
provider: google
|
|
1068
1180
|
display_name: Gemini 2.5 Flash
|
|
1069
1181
|
description: Legacy Flash model for multimodal reasoning and tool use; available to previous users.
|
|
1070
|
-
notes:
|
|
1071
|
-
Audio input is $1 and cached audio $0.10 per 1M tokens, not represented by these text rates.'
|
|
1182
|
+
notes: "Shuts down December 1, 2026; requests then route to gemini-3.5-flash-lite (Gemini deprecations page, checked 2026-11-20). Not selectable in AI Studio; API access restricted to existing users since 2026-09-18. Paid Standard text/image/video rates. Audio input is $1 and cached audio $0.10 per 1M tokens, not represented by these text rates."
|
|
1072
1183
|
input_price: 0.3
|
|
1073
1184
|
output_price: 2.5
|
|
1074
1185
|
cached_input_price: 0.03
|
|
@@ -1102,7 +1213,7 @@ models:
|
|
|
1102
1213
|
display_name: Gemini 2.5 Flash TTS
|
|
1103
1214
|
description: Text-to-speech variant of Gemini 2.5 Flash for audio generation at lower cost.
|
|
1104
1215
|
notes: 'Native Gemini TTS API required; not supported by the Agent chat route. Output is audio at $10 per 1M audio
|
|
1105
|
-
tokens, not a text output rate. 16,384 max output tokens. Shuts down 2026-
|
|
1216
|
+
tokens, not a text output rate. 16,384 max output tokens. Shuts down 2026-12-18; migrate to
|
|
1106
1217
|
gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts. https://ai.google.dev/gemini-api/docs/deprecations'
|
|
1107
1218
|
input_price: 0.5
|
|
1108
1219
|
output_price: null
|
|
@@ -1112,8 +1223,7 @@ models:
|
|
|
1112
1223
|
provider: xai
|
|
1113
1224
|
display_name: Grok Build 0.1
|
|
1114
1225
|
description: Coding model for agentic software engineering with image input, reasoning, and function calling.
|
|
1115
|
-
notes: 'Global API rates. At >=200K prompt tokens, full-request input/output/cache rates double.
|
|
1116
|
-
Batch API is not supported. Aliases include grok-code-fast-1; use this canonical ID.'
|
|
1226
|
+
notes: "Not listed on docs.x.ai/docs/models (checked 2026-11-20) and the live API returns 'does not exist or your team does not have access'. Treat as unavailable unless your team has access. Global API rates. At >=200K prompt tokens, full-request input/output/cache rates double. Batch API is not supported. Aliases include grok-code-fast-1; use this canonical ID."
|
|
1117
1227
|
input_price: 1.0
|
|
1118
1228
|
output_price: 2.0
|
|
1119
1229
|
cached_input_price: 0.2
|
|
@@ -1197,7 +1307,7 @@ models:
|
|
|
1197
1307
|
provider: groq
|
|
1198
1308
|
display_name: Qwen3.8 27B (Groq)
|
|
1199
1309
|
description: Qwen3.8 27B multimodal with hybrid thinking/instruct modes, delivered at Groq's ultra-low latency.
|
|
1200
|
-
notes: Preview tier. Qwen3.6 27B was retired for free/developer accounts on 2026-09-14. See https://console.groq.com/docs/deprecations
|
|
1310
|
+
notes: "Preview tier; Groq may discontinue preview models at short notice (checked 2026-11-20). Qwen3.6 27B was retired for free/developer accounts on 2026-09-14. See https://console.groq.com/docs/deprecations"
|
|
1201
1311
|
input_price: 0.8
|
|
1202
1312
|
output_price: 4.0
|
|
1203
1313
|
context_window: 131072
|
|
@@ -1738,7 +1848,7 @@ models:
|
|
|
1738
1848
|
provider: moonshot
|
|
1739
1849
|
display_name: Kimi K2.7 Code
|
|
1740
1850
|
description: Moonshot coding-focused multimodal model with thinking mode for long-context programming agents.
|
|
1741
|
-
notes: Cache-hit input is $0.19 / 1M tokens.
|
|
1851
|
+
notes: "Moonshot lists this under Deprecated Models (checked 2026-11-20): unavailable to newly registered users; users registered before November 1 can keep using it; deprecation expected late December 2026. K2-series models retire by end of December 2026. Cache-hit input is $0.19 / 1M tokens."
|
|
1742
1852
|
input_price: 0.95
|
|
1743
1853
|
output_price: 4.0
|
|
1744
1854
|
cached_input_price: 0.19
|
|
@@ -1748,7 +1858,7 @@ models:
|
|
|
1748
1858
|
provider: moonshot
|
|
1749
1859
|
display_name: Kimi K2.7 Code HighSpeed
|
|
1750
1860
|
description: Same Kimi K2.7 Code weights with higher output throughput (~180 tok/s) for interactive coding agents.
|
|
1751
|
-
notes: Cache-hit input is $0.38 / 1M tokens.
|
|
1861
|
+
notes: "Moonshot lists this under Deprecated Models (checked 2026-11-20): unavailable to newly registered users; users registered before November 1 can keep using it; deprecation expected late December 2026. K2-series models retire by end of December 2026. Cache-hit input is $0.38 / 1M tokens."
|
|
1752
1862
|
input_price: 1.9
|
|
1753
1863
|
output_price: 8.0
|
|
1754
1864
|
cached_input_price: 0.38
|
|
@@ -1758,7 +1868,7 @@ models:
|
|
|
1758
1868
|
provider: moonshot
|
|
1759
1869
|
display_name: Kimi K2.6
|
|
1760
1870
|
description: Moonshot general-purpose multimodal MoE with thinking/non-thinking modes for chat, agents, and vision.
|
|
1761
|
-
notes: Cache-hit input is $0.16 / 1M tokens.
|
|
1871
|
+
notes: "Moonshot lists this under Deprecated Models (checked 2026-11-20): unavailable to newly registered users; users registered before November 1 can keep using it; deprecation expected late December 2026. K2-series models retire by end of December 2026. Cache-hit input is $0.16 / 1M tokens."
|
|
1762
1872
|
input_price: 0.95
|
|
1763
1873
|
output_price: 4.0
|
|
1764
1874
|
cached_input_price: 0.16
|