modelmix 5.2.3 → 5.3.1

package/README.md

@@ -138,11 +138,11 @@ const ETH = ModelMix.new()
 console.log(ETH.price);
 ```
 
-**This example uses providers with free quotas (Groq, Cerebras, and Together). OpenRouter is disabled because its GPT-OSS 120B route is no longer free. If one model runs out of quota, ModelMix automatically falls back to the next model in the chain.**
+**This example chains two multi-provider shortcuts. Each one registers its default provider and accepts additional providers through `mix`; if one model fails, ModelMix automatically falls back to the next model in the chain.**
 ```javascript
 ModelMix.new()
-  .gptOss()
-  .kimiK25()
+  .kimiK26()
+  .GLM52()
   .addText('What is the capital of France?');
 ```
 
@@ -160,7 +160,9 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | Method | Provider | Model | Input / 1M | Output / 1M |
 | --- | --- | --- | ---: | ---: |
 | `gpt6astra()` | OpenAI | gpt-6-astra | [\$10.00][1] | [\$50.00][1] |
-| `gpt56sol()` | OpenAI | gpt-5.6-sol | [\$5.00][1] | [\$30.00][1] |
+| `gpt6sol()` | OpenAI | gpt-6-sol | [\$2.00][1] | [\$10.00][1] |
+| `gpt6luna()` | OpenAI | gpt-6-luna | [\$0.10][1] | [\$0.50][1] |
+| `gpt56sol()` | OpenAI | gpt-5.6-sol | [\$4.00][1] | [\$20.00][1] |
 | `gpt56terra()` | OpenAI | gpt-5.6-terra | [\$2.00][1] | [\$12.00][1] |
 | `gpt56luna()` | OpenAI | gpt-5.6-luna | [\$0.20][1] | [\$1.20][1] |
 | `gpt55()` | OpenAI | gpt-5.5 | [\$5.00][1] | [\$30.00][1] |
@@ -169,14 +171,9 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `gpt54mini()` | OpenAI | gpt-5.4-mini | [\$0.75][1] | [\$4.50][1] |
 | `gpt54nano()` | OpenAI | gpt-5.4-nano | [\$0.20][1] | [\$1.25][1] |
 | `gpt54pro()` | OpenAI | gpt-5.4-pro | [\$30.00][1] | [\$180.00][1] |
-| `gpt53codex()` | OpenAI | gpt-5.3-codex | [\$1.75][1] | [\$14.00][1] |
-| `gpt53chat()` | OpenAI | gpt-5.3-chat-latest | [\$1.75][1] | [\$14.00][1] |
 | `gpt52()` | OpenAI | gpt-5.2 | [\$1.75][1] | [\$14.00][1] |
-| `gpt51()` | OpenAI | gpt-5.1 | [\$1.25][1] | [\$10.00][1] |
-| `gpt5()` | OpenAI | gpt-5 | [\$1.25][1] | [\$10.00][1] |
 | `gpt5mini()` | OpenAI | gpt-5-mini | [\$0.25][1] | [\$2.00][1] |
 | `gpt5nano()` | OpenAI | gpt-5-nano | [\$0.05][1] | [\$0.40][1] |
-| `gptOss()` | Multi-provider | gpt-oss-120B | [\$0.15][7] | [\$0.60][7] |
 | `fable51()` | Anthropic | claude-fable-5-1 | [\$10.00][2] | [\$50.00][2] |
 | `fable5()` | Anthropic | claude-fable-5 | [\$10.00][2] | [\$50.00][2] |
 | `opus55()` | Anthropic | claude-opus-5-5 | [\$5.00][2] | [\$25.00][2] |
@@ -185,9 +182,7 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `opus47()` | Anthropic | claude-opus-4-7 | [\$5.00][2] | [\$25.00][2] |
 | `opus46()` | Anthropic | claude-opus-4-6 | [\$5.00][2] | [\$25.00][2] |
 | `sonnet5()` | Anthropic | claude-sonnet-5 | [\$3.00][2] | [\$15.00][2] |
-| `sonnet46()` | Anthropic | claude-sonnet-4-6 | [\$3.00][2] | [\$15.00][2] |
 | `haiku45()` | Anthropic | claude-haiku-4-5-20251001 | [\$1.00][2] | [\$5.00][2] |
-| `gemini31pro()` | Google | gemini-3.1-pro-preview | [\$2.00][3] | [\$12.00][3] |
 | `gemini38flash()` | Google | gemini-3.8-flash | [\$0.75][3] | [\$3.75][3] |
 | `gemini37flash()` | Google | gemini-3.7-flash | [\$0.75][3] | [\$3.75][3] |
 | `gemini36flash()` | Google | gemini-3.6-flash | [\$0.75][3] | [\$3.75][3] |
@@ -196,10 +191,7 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `gemini31flashLite()` | Google | gemini-3.1-flash-lite-preview | [\$0.25][3] | [\$1.50][3] |
 | `grok47()` | Grok | grok-4.7 | — | — |
 | `grok46()` | Grok | grok-4.6 | [\$2.00][6] | [\$6.00][6] |
-| `grok45()` | Grok | grok-4.5 | [\$2.00][6] | [\$6.00][6] |
 | `grok43()` | Grok | grok-4.3 | [\$1.25][6] | [\$2.50][6] |
-| `grok420multiAgent()` | Grok | grok-4.20-multi-agent-0309 | [\$1.25][6] | [\$2.50][6] |
-| `grok420()` | Grok | grok-4.20-0309 (†) | [\$1.25][6] | [\$2.50][6] |
 | `museGlimmer30b()` | Fireworks | models/muse-glimmer-30b | [\$0.35][17] | [\$1.50][17] |
 | `museSpark12()` | OpenRouter | meta/muse-spark-1.2 | [\$1.25][25] | [\$4.25][25] |
 | `museSpark12c()` | OpenRouter | meta/muse-spark-1.2-contributor | [\$0.10][22] | [\$0.20][22] |
@@ -212,7 +204,7 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `qwen3827b()` | OpenRouter | qwen/qwen3.8-27b | [\$0.45][15] | [\$3.20][15] |
 | `qwen38flash()` | OpenRouter | qwen/qwen3.8-flash | [\$0.16][19] | [\$0.47][19] |
 | `deepseekV4Flash()` | Fireworks | models/deepseek-v4-flash | [\$0.14][10] | [\$0.28][10] |
-| `deepseekV41Flash()` | OpenRouter | deepseek/deepseek-v4.1-flash | [\$0.15][27] | [\$0.60][27] |
+| `deepseekV41Flash()` | DeepSeek | deepseek-flash | [\$0.30][29] | [\$1.20][29] |
 | `deepseekV4Pro()` | Fireworks | models/deepseek-v4-pro-0813 | [\$1.32][12] | [\$3.96][12] |
 | `deepseekPro()` | OpenRouter | deepseek/deepseek-v4-pro-0813 | [\$0.5808][28] | [\$1.7424][28] |
 | `GLM53()` | OpenRouter | z-ai/glm-5.3 | [\$1.40][16] | [\$4.40][16] |
@@ -228,7 +220,6 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `hermes3()` | Lambda | Hermes-3-Llama-3.1-405B-FP8 | [\$0.80][8] | [\$0.80][8] |
 | `kimiK3()` | Moonshot | kimi-k3 | [\$3.00][11] | [\$15.00][11] |
 | `kimiK27Code()` | Together | Kimi-K2.7-Code | [\$0.95][7] | [\$4.00][7] |
-| `kimiK25()` | Together | Kimi-K2.5 | [\$0.50][7] | [\$2.80][7] |
 | `kimiK26()` | Fireworks | models/kimi-k2p6 | [\$0.95][10] | [\$4.00][10] |
 
 Gemini 3.8 Flash, 3.7 Flash, and 3.6 Flash use Google's introductory standard pricing through December 31, 2026; standard rates double on January 1, 2027.
@@ -237,9 +228,9 @@ Gemini 3.8 Flash, 3.7 Flash, and 3.6 Flash use Google's introductory standard pr
 
 `fable51()` uses the official Anthropic API by default (`claude-fable-5-1`). Pass `mix: { openrouter: true }` to append [`anthropic/claude-fable-5.1`][21] as its fallback.
 
-Every textual GPT-5 and GPT-6 shortcut in the table uses the official OpenAI API by default. Pass `mix: { openrouter: true }` to `ModelMix.new()` or to an individual shortcut to append the matching [`openai/*` OpenRouter route][23] as its fallback. `gpt53chat()` maps the official `gpt-5.3-chat-latest` alias to `openai/gpt-5.3-chat`. Realtime shortcuts remain official-only because they use OpenAI's WebSocket transport.
+Every textual GPT-5 and GPT-6 shortcut in the table uses the official OpenAI API by default. Pass `mix: { openrouter: true }` to `ModelMix.new()` or to an individual shortcut to append the matching [`openai/*` OpenRouter route][23] as its fallback. Realtime shortcuts remain official-only because they use OpenAI's WebSocket transport.
 
-OpenRouter fallbacks are disabled globally by default and are appended only with `mix.openrouter: true`. Shortcuts whose primary provider is OpenRouter, such as `qwen36plus()`, are unaffected. The multi-provider shortcuts also expose the current catalog alternatives: `gptOss()` supports NVIDIA and Fireworks; `qwen37plus()` supports Together; `kimiK27Code()` supports Fireworks and OpenRouter; `kimiK3()` supports Fireworks, OpenRouter, and Together; `GLM52()` supports Fireworks and OpenRouter; and both MiniMax shortcuts support Fireworks. `minimaxM27()` keeps every explicitly enabled provider in its fallback chain.
+OpenRouter fallbacks are disabled globally by default and are appended only with `mix.openrouter: true`. Shortcuts whose primary provider is OpenRouter, such as `qwen36plus()`, are unaffected. The multi-provider shortcuts also expose the current catalog alternatives: `qwen37plus()` supports Together; `kimiK27Code()` supports Fireworks and OpenRouter; `kimiK3()` supports Fireworks, OpenRouter, and Together; `GLM52()` supports Fireworks and OpenRouter; and both MiniMax shortcuts support Fireworks. `minimaxM27()` keeps every explicitly enabled provider in its fallback chain.
 
 [1]: https://platform.openai.com/docs/pricing "Pricing | OpenAI"
 [2]: https://docs.anthropic.com/en/docs/about-claude/pricing "Pricing - Anthropic"
@@ -268,16 +259,14 @@ OpenRouter fallbacks are disabled globally by default and are appended only with
 
 [25]: https://openrouter.ai/meta/muse-spark-1.2 "Muse Spark 1.2 on OpenRouter"
 [26]: https://openrouter.ai/meta/muse-spark-1.3-contributor "Muse Spark 1.3 Contributor on OpenRouter"
-[27]: https://openrouter.ai/deepseek/deepseek-v4.1-flash "DeepSeek V4.1 Flash on OpenRouter"
 [28]: https://openrouter.ai/deepseek/deepseek-v4-pro-0813 "DeepSeek V4 Pro 0813 on OpenRouter"
+[29]: https://api-docs.deepseek.com/quick_start/pricing/ "DeepSeek API Pricing"
 
 `deepseekPro()` uses the pinned OpenRouter model `deepseek/deepseek-v4-pro-0813` and requires `OPENROUTER_API_KEY`. Use `chain('deepseekPro@100')` for maximum reasoning effort. Cost estimates use the listed base rates and $0.05808/M cached input tokens; actual rates may change, including provider and time-based pricing.
 
-`deepseekV41Flash()` supports text and image input through OpenRouter and requires `OPENROUTER_API_KEY`. Use `chain('deepseekV41Flash@100')` for maximum reasoning effort. Cost estimates use the listed base rates and $0.015/M cached input tokens; actual OpenRouter pricing varies by provider and time.
+By default `deepseekV41Flash()` uses the native DeepSeek API (`https://api.deepseek.com/chat/completions`, model `deepseek-flash`, currently V4.1 Flash) and requires `DEEPSEEK_API_KEY`; `MixDeepSeek` remains available for explicit `.attach()` calls. Use `chain('deepseekV41Flash@100')` for maximum reasoning effort. Native cost estimates use [peak rates](https://api-docs.deepseek.com/quick_start/pricing/): $0.30 input / $0.006 cached input / $1.20 output per 1M tokens; actual off-peak charges are half.
 
-For direct Fireworks access, use `deepseekV41Flash({ mix: { fireworks: true, openrouter: false } })` with `FIREWORKS_API_KEY`. This selects `accounts/fireworks/models/deepseek-v4p1-flash`, priced at [$0.22 input / $0.007 cached input / $0.66 output per 1M tokens](https://fireworks.ai/models/deepseek-ai/deepseek-v4p1-flash). Set both providers to `true` to try Fireworks first and fall back to OpenRouter.
-
-For the native DeepSeek API, use `deepseekV41Flash({ mix: { deepseek: true, openrouter: false } })` with `DEEPSEEK_API_KEY`. It calls `https://api.deepseek.com/chat/completions` with `deepseek-flash`, currently DeepSeek V4.1 Flash. `MixDeepSeek` is also available for explicit `.attach()` calls. Native cost estimates use [peak rates](https://api-docs.deepseek.com/quick_start/pricing/): $0.30 input / $0.006 cached input / $1.20 output per 1M tokens; actual off-peak charges are half. When all three providers are enabled, the order is DeepSeek → Fireworks → OpenRouter.
+Optional routes: `mix: { fireworks: true }` selects `accounts/fireworks/models/deepseek-v4p1-flash` with `FIREWORKS_API_KEY`, priced at [$0.22 input / $0.007 cached input / $0.66 output per 1M tokens](https://fireworks.ai/models/deepseek-ai/deepseek-v4p1-flash), and `mix: { openrouter: true }` selects `deepseek/deepseek-v4.1-flash` with `OPENROUTER_API_KEY`, which accepts text and image input. With all three providers enabled the order is DeepSeek → Fireworks → OpenRouter.
 
 Muse Spark methods ending in `c` select Contributor: prompts and outputs may be used to improve Meta products. Methods without `c` select the standard tier. `museSpark12()` now selects standard; use `museSpark12c()` for the previous Contributor behavior.
 
@@ -320,6 +309,7 @@ ModelMix.new().effort(-1).minimaxM3().addText('...').message();
 
 - **Gemini:** Gemini 3+ uses bands 0–24 / 25–49 / 50–74 / 75–100. Gemini 3.8 Flash and 3.7 Flash clamp these bands to `low` / `low` / `medium` / `high`; `-1` leaves their native `medium` default unchanged. Gemini 2.5 maps 0–100 to `thinkingBudget`.
 - **GPT-6 Astra:** 0–39 maps to `low`, 40–59 to `medium`, 60–79 to `high`, 80–99 to `xhigh`, and 100 to `max`. [Model details](https://developers.openai.com/api/docs/models/gpt-6-astra). Cache reads cost $1.00 and cache writes $12.50 per 1M tokens; requests over 272K input tokens apply 2× input/cache and 1.5× output rates.
+- **GPT-6 Sol and GPT-6 Luna:** `100` maps to `max`; the remaining bands match OpenAI, so 0–19 is `none` and 80–99 is `xhigh`. [Sol model details](https://developers.openai.com/api/docs/models/gpt-6-sol), [Luna model details](https://developers.openai.com/api/docs/models/gpt-6-luna). Cache reads cost $0.20 and $0.01 and cache writes $2.50 and $0.125 per 1M tokens; requests over 272K input tokens apply the same 2× input/cache and 1.5× output rates as Astra.
 - **GPT-5.6:** `100` maps to `max`; 80–99 remains `xhigh`.
 - **Qwen 3.8 27B and Flash:** 0–39 / 40–79 / 80–100 map to `low` / `medium` / `xhigh`; `-1` leaves the native `xhigh` default unchanged. Qwen 3.8 Flash is the managed production version based on the open-weight Flash-Next architecture.
 - **GLM 5.3 and GLM 5.3 Flash:** reasoning is mandatory; 0–39 / 40–79 / 80–100 map to `low` / `high` / `max`; `-1` leaves the native `max` default unchanged.
@@ -334,8 +324,7 @@ ModelMix.new().effort(-1).minimaxM3().addText('...').message();
 
 The former `*think()` methods were removed. Use `.effort(n).<model>()` with `0`–`100` or `-1` instead.
 
-- **Kimi:** use `kimiK25()` or `kimiK26()`.
-- **Grok 4.20:** `.grok420()` selects the non-reasoning model. Use `.effort(20+).grok420()` or `.effort(-1).grok420()` to select the reasoning model.
+- **Kimi:** use `kimiK26()`.
 
 ## 🔄 Templates
 
@@ -831,9 +820,9 @@ await ask('Summarize support ticket 456.');
 
 The contents of `support.md` and the cache key stay the same between calls; only the final question changes. The first request may report `cacheWrite > 0`, while later requests confirm reuse with `cached > 0`. For GPT-5.6, the stable prefix must contain at least 1,024 tokens. Keep all variable content after the breakpoint, and change `prompt_cache_key` when the stable instructions change.
 
-### GPT-5.6 prompt caching
+### GPT-5.6 and GPT-6 prompt caching
 
-GPT-5.6 supports implicit or explicit caching through `prompt_cache_options`. Put the explicit breakpoint at the end of the stable prefix; the provider only caches prompts with at least 1,024 tokens.
+GPT-5.6 and the GPT-6 models (Astra, Sol, and Luna) support implicit or explicit caching through `prompt_cache_options`. Put the explicit breakpoint at the end of the stable prefix; the provider only caches prompts with at least 1,024 tokens.
 
 ```javascript
 const model = ModelMix.new()
@@ -854,16 +843,18 @@ console.log(result.tokens.cached, result.tokens.cacheWrite, result.tokens.cost);
 
 The provider-neutral `cache: { breakpoint: true }` option is accepted by `addTextFromFile()`, `addImage()`, `addImageFromUrl()`, and `addImageFromBuffer()`. Responses-native `input_text`, `input_image`, and `input_file` blocks preserve the native `prompt_cache_breakpoint` field when supplied directly through `options.messages`.
 
-GPT-5.6 uses `prompt_cache_options.ttl`; `prompt_cache_retention` remains available for earlier OpenAI models. ModelMix rejects the incompatible control instead of silently dropping it. For GPT-5.6 requests over 272K input tokens, the cost calculation applies the documented 2× input and 1.5× output multipliers to the complete request, including cache reads and writes.
+GPT-5.6 and GPT-6 use `prompt_cache_options.ttl`; `prompt_cache_retention` remains available for earlier OpenAI models. ModelMix rejects the incompatible control instead of silently dropping it. For GPT-5.6 requests over 272K input tokens, the cost calculation applies the documented 2× input and 1.5× output multipliers to the complete request, including cache reads and writes.
 
 GPT-5.6 prices per 1M tokens:
 
 | Model | Input | Cached input | Cache write | Output |
 | --- | ---: | ---: | ---: | ---: |
-| `gpt-5.6-sol` | $5.00 | $0.50 | $6.25 | $30.00 |
+| `gpt-5.6-sol` | $4.00 | $0.40 | $5.00 | $20.00 |
 | `gpt-5.6-terra` | $2.00 | $0.20 | $2.50 | $12.00 |
 | `gpt-5.6-luna` | $0.20 | $0.02 | $0.25 | $1.20 |
 
+GPT-5.6 Sol's $4.00 input and $20.00 output rates are promotional and available at least through November 21, 2026.
+
 ### Cross-provider cache fallback
 
 Neutral breakpoints are translated at the last moment by each provider adapter. Native request policies remain scoped to their model, so they cannot leak into a fallback request:
@@ -885,7 +876,7 @@ const model = ModelMix.new()
   .addText('Answer this variable request.');
 ```
 
-GPT-5.6 receives `prompt_cache_breakpoint`; Anthropic receives `cache_control`; older OpenAI models and providers without an equivalent omit the marker. When a neutral explicit breakpoint is present for Anthropic, its model-scoped `cache_control` becomes that block's policy instead of adding an automatic breakpoint after the variable suffix.
+GPT-5.6 and GPT-6 receive `prompt_cache_breakpoint`; Anthropic receives `cache_control`; older OpenAI models and providers without an equivalent omit the marker. When a neutral explicit breakpoint is present for Anthropic, its model-scoped `cache_control` becomes that block's policy instead of adding an automatic breakpoint after the variable suffix.
 
 ## 🔧 Model Context Protocol (MCP) Integration
 
@@ -1055,7 +1046,7 @@ Pass `mix` to `benchmark()` to select providers for the criteria model, particip
 
 JSON parsing accepts a Markdown code block or a lone closing triple-backtick delimiter after valid JSON. Other extra content, malformed JSON, and invalid evaluation scores remain errors.
 
-Run `node demo/benchmark.js` for a complete five-model comparison. It allows up to 32,768 output tokens per call, prints intermediate progress, a final ranking and error details, then saves each response as Markdown together with the full JSON report under `demo/results/`.
+Run `node demo/benchmark.js` for a complete model comparison. It allows up to 32,768 output tokens per call, prints intermediate progress, a final ranking and error details, then saves each response as Markdown together with the full JSON report under `demo/results/`.
 
 ### Recursive Language Model plugin
 

package/demo/benchmark.js

@@ -11,16 +11,13 @@ const { benchmark } = benchmarkPlugin;
 const task = await readFile(process.argv[2] || new URL('./prompts/story.txt', import.meta.url), 'utf8');
 
 const models = [
-    // 'opus50@20',
-    'gpt6astra@0',
-    'gpt6astra@20',
+    'gpt6sol@40',
+    'gpt6luna@40',
     'gpt56sol@40',
-    // 'gemini38flash@20',
-    // 'grok46@20',
-    'deepseekV41Flash@60'
+    'gpt56luna@40'
 ];
 
-console.log(`Running benchmark: ${models.length} models, 26 sequential model calls.`);
+console.log(`Running benchmark: ${models.length} models, 17 sequential model calls.`);
 
 const report = await ModelMix.new({
     config: { debug: 1 },

package/demo/demo.js

@@ -24,9 +24,9 @@ const pplxSettings = {
 
 mmix.assign({ name: 'ALF' });
 
-console.log("\n" + '--------| gpt51() |--------');
+console.log("\n" + '--------| gpt52() |--------');
 const gptArgs = { options: { reasoning_effort: "none", verbosity: "low" } };
-const gpt = mmix.gpt51(gptArgs).addText("Have you ever eaten a <%- animal %>?");
+const gpt = mmix.gpt52(gptArgs).addText("Have you ever eaten a <%- animal %>?");
 gpt.assignKey('animal', 'cat');
 await gpt.json({ time: '24:00:00', message: 'Hello' }, { time: 'Time in format HH:MM:SS' });
 

package/demo/fallback.js

@@ -15,7 +15,7 @@ const mmix = new ModelMix({
     }
 });
 
-mmix.sonnet46({ config: { url: 'fail' } }).gpt5nano();
+mmix.sonnet5({ config: { url: 'fail' } }).gpt5nano();
 
 async function main() {
     mmix.addText('hola, como estas?');

package/demo/free.js

@@ -2,9 +2,9 @@ import { ModelMix } from '../index.js';
 try { process.loadEnvFile(); } catch {}
 
 const ai = ModelMix.new({ config: { debug: 2 } })
-    .gptOss()
-    .kimiK25()
+    .kimiK26()
+    .GLM52()
     .addText('What is the capital of France?');
 
 const response = await ai.message();
-console.log('Response from a free-tier provider:', response);
+console.log('Response from chained multi-provider models:', response);

package/demo/gemini.js

@@ -20,9 +20,9 @@ const flash = await mmix.gemini38flash()
 
 console.log(flash);
 
-// Using Gemini 3.1 Pro with custom config
-console.log("\n" + '--------| gemini31pro() with JSON response |--------');
-const pro = mmix.new().gemini31pro();
+// Using Gemini 3.7 Flash with custom config
+console.log("\n" + '--------| gemini37flash() with JSON response |--------');
+const pro = mmix.new().gemini37flash();
 
 pro.addText('Give me a fun fact about cats');
 

package/demo/json.js

@@ -3,7 +3,6 @@ try { process.loadEnvFile(); } catch {}
 
 const model = await ModelMix.new({ options: { max_tokens: 10000 }, config: { debug: 3 } })
     .gpt54nano()
-    // .gptOss()
     // .o4mini()
     // .gpt45()
     // .gemini37flash()

package/demo/round-robin.js

@@ -11,11 +11,11 @@ const ai = ModelMix.new({
     },
     mix: { openrouter: false } // Exclude OpenRouter (free tier often rate-limited)
 })
-    .gptOss();
+    .kimiK26({ mix: { fireworks: true, together: true } });
 
 console.log('Making 6 requests with round robin enabled...\n');
 
-// Make 6 requests to see rotation through all models (cerebras + groq)
+// Make 6 requests to see rotation through all models (fireworks + together)
 for (let i = 1; i <= 6; i++) {
     const result = await ai.new()
         .addText(`Calculate ${i} * 2`)

package/demo/short.js

@@ -10,7 +10,7 @@ const setup = {
 
 const mmix = await ModelMix.new(setup)
     .fable51({ mix: { openrouter: true } }) // (main + provider fallback) Anthropic/OpenRouter Claude Fable 5.1
-    .sonnet46() // (fallback 2) Anthropic claude-sonnet-4-6
+    .sonnet5() // (fallback 2) Anthropic claude-sonnet-5
     .gpt56luna({ mix: { openrouter: true } }) // (fallback 3 + provider fallback) OpenAI/OpenRouter gpt-5.6-luna
     .gemini38flash() // (fallback 4) Google gemini-3.8-flash
     .gpt5nano({ mix: { openrouter: true } }) // (fallback 5 + provider fallback) OpenAI/OpenRouter gpt-5-nano

package/effort.js

@@ -43,6 +43,8 @@ const GLM53_BANDS = [
 /** Exact model → supported OpenAI reasoning_effort values */
 const OPENAI_MODEL_LEVELS = {
     'gpt-6-astra': ['low', 'medium', 'high', 'xhigh', 'max'],
+    'gpt-6-sol': ['none', 'low', 'medium', 'high', 'xhigh', 'max'],
+    'gpt-6-luna': ['none', 'low', 'medium', 'high', 'xhigh', 'max'],
     'gpt-5.6-sol': ['none', 'low', 'medium', 'high', 'xhigh', 'max'],
     'gpt-5.6-terra': ['none', 'low', 'medium', 'high', 'xhigh', 'max'],
     'gpt-5.6-luna': ['none', 'low', 'medium', 'high', 'xhigh', 'max'],
@@ -139,7 +141,7 @@ const PROVIDER_FAMILY_BY_CLASS = {
     MixCustom: null,
 };
 
-/** Logical alias from `.grok420()` — resolved to reasoning / non-reasoning at request time. */
+/** Logical alias key `grok-4.20-0309` — resolved to reasoning / non-reasoning at request time. */
 const GROK420_ALIAS = 'grok-4.20-0309';
 const GROK420_REASONING = 'grok-4.20-0309-reasoning';
 const GROK420_NON_REASONING = 'grok-4.20-0309-non-reasoning';

package/index.d.ts

@@ -445,10 +445,8 @@ export declare class ModelMix {
   attach(key: string, provider: MixCustom): this;
 
   // OpenAI
-  gpt5(args?: ModelAttachArgs): this;
   gpt5mini(args?: ModelAttachArgs): this;
   gpt5nano(args?: ModelAttachArgs): this;
-  gpt51(args?: ModelAttachArgs): this;
   gpt52(args?: ModelAttachArgs): this;
   gpt54(args?: ModelAttachArgs): this;
   gpt54mini(args?: ModelAttachArgs): this;
@@ -457,14 +455,13 @@ export declare class ModelMix {
   gpt55(args?: ModelAttachArgs): this;
   gpt55pro(args?: ModelAttachArgs): this;
   gpt6astra(args?: ModelAttachArgs): this;
+  gpt6sol(args?: ModelAttachArgs): this;
+  gpt6luna(args?: ModelAttachArgs): this;
   gpt56sol(args?: ModelAttachArgs): this;
   gpt56terra(args?: ModelAttachArgs): this;
   gpt56luna(args?: ModelAttachArgs): this;
   gptRealtime(args?: ModelAttachArgs): this;
   gptRealtimeMini(args?: ModelAttachArgs): this;
-  gpt53codex(args?: ModelAttachArgs): this;
-  gpt53chat(args?: ModelAttachArgs): this;
-  gptOss(args?: ModelAttachArgs): this;
 
   // Anthropic
   fable51(args?: ModelAttachArgs): this;
@@ -478,12 +475,10 @@ export declare class ModelMix {
   opus46(args?: ModelAttachArgs): this;
   sonnet50(args?: ModelAttachArgs): this;
   sonnet5(args?: ModelAttachArgs): this;
-  sonnet46(args?: ModelAttachArgs): this;
   sonnet45(args?: ModelAttachArgs): this;
   haiku45(args?: ModelAttachArgs): this;
 
   // Google
-  gemini31pro(args?: ModelAttachArgs): this;
   gemini38flash(args?: ModelAttachArgs): this;
   gemini37flash(args?: ModelAttachArgs): this;
   gemini36flash(args?: ModelAttachArgs): this;
@@ -498,11 +493,7 @@ export declare class ModelMix {
   // Grok
   grok47(args?: ModelAttachArgs): this;
   grok46(args?: ModelAttachArgs): this;
-  grok45(args?: ModelAttachArgs): this;
   grok43(args?: ModelAttachArgs): this;
-  grok420multiAgent(args?: ModelAttachArgs): this;
-  /** Non-reasoning by default; with `.effort(20+)` / `-1` uses the reasoning model at request time. */
-  grok420(args?: ModelAttachArgs): this;
 
   // Multi-provider
   museGlimmer30b(args?: ModelAttachArgs): this;
@@ -522,7 +513,6 @@ export declare class ModelMix {
   kimiK26(args?: ModelAttachArgs): this;
   kimiK27Code(args?: ModelAttachArgs): this;
   kimiK3(args?: ModelAttachArgs): this;
-  kimiK25(args?: ModelAttachArgs): this;
   lmstudio(model?: string, args?: ModelAttachArgs): this;
   minimaxM27(args?: ModelAttachArgs): this;
   minimaxM3(args?: ModelAttachArgs): this;

package/index.js

@@ -463,18 +463,12 @@ class ModelMix {
         return this;
     }
 
-    gpt5(args = {}) {
-        return this._attachOpenAIWithOpenRouter('gpt-5', MixOpenAI, args);
-    }
     gpt5mini(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-5-mini', MixOpenAI, args);
     }
     gpt5nano(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-5-nano', MixOpenAI, args);
     }
-    gpt51(args = {}) {
-        return this._attachOpenAIWithOpenRouter('gpt-5.1', MixOpenAIResponses, args);
-    }
     gpt52(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-5.2', MixOpenAIResponses, args);
     }
@@ -499,6 +493,12 @@ class ModelMix {
     gpt6astra(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-6-astra', MixOpenAIResponses, args);
     }
+    gpt6sol(args = {}) {
+        return this._attachOpenAIWithOpenRouter('gpt-6-sol', MixOpenAIResponses, args);
+    }
+    gpt6luna(args = {}) {
+        return this._attachOpenAIWithOpenRouter('gpt-6-luna', MixOpenAIResponses, args);
+    }
     gpt56sol(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-5.6-sol', MixOpenAIResponses, args);
     }
@@ -514,25 +514,6 @@ class ModelMix {
     gptRealtimeMini({ options = {}, config = {} } = {}) {
         return this.attach('gpt-realtime-mini', new MixOpenAIWebSocket({ options, config }));
     }
-    gpt53codex(args = {}) {
-        return this._attachOpenAIWithOpenRouter('gpt-5.3-codex', MixOpenAIResponses, args);
-    }
-    gpt53chat(args = {}) {
-        return this._attachOpenAIWithOpenRouter('gpt-5.3-chat-latest', MixOpenAIResponses, {
-            ...args,
-            openRouterKey: 'openai/gpt-5.3-chat'
-        });
-    }
-    gptOss({ options = {}, config = {}, mix = {} } = {}) {
-        mix = { ...this.mix, ...mix };
-        if (mix.nvidia) this.attach('openai/gpt-oss-120b', new MixNVIDIA({ options, config }));
-        if (mix.fireworks) this.attach('accounts/fireworks/models/gpt-oss-120b', new MixFireworks({ options, config }));
-        if (mix.together) this.attach('openai/gpt-oss-120b', new MixTogether({ options, config }));
-        if (mix.cerebras) this.attach('gpt-oss-120b', new MixCerebras({ options, config }));
-        if (mix.groq) this.attach('openai/gpt-oss-120b', new MixGroq({ options, config }));
-        if (mix.openrouter) this.attach('openai/gpt-oss-120b', new MixOpenRouter({ options, config }));
-        return this;
-    }
     fable50({ options = {}, config = {} } = {}) {
         return this.attach('claude-fable-5', new MixAnthropic({ options, config }));
     }
@@ -569,18 +550,12 @@ class ModelMix {
     sonnet5(args = {}) {
         return this.sonnet50(args);
     }
-    sonnet46({ options = {}, config = {} } = {}) {
-        return this.attach('claude-sonnet-4-6', new MixAnthropic({ options, config }));
-    }
     sonnet45({ options = {}, config = {} } = {}) {
         return this.attach('claude-sonnet-4-5-20250929', new MixAnthropic({ options, config }));
     }
     haiku45({ options = {}, config = {} } = {}) {
         return this.attach('claude-haiku-4-5-20251001', new MixAnthropic({ options, config }));
     }
-    gemini31pro({ options = {}, config = {} } = {}) {
-        return this.attach('gemini-3.1-pro-preview', new MixGoogle({ options, config }));
-    }    
     gemini38flash({ options = {}, config = {} } = {}) {
         return this.attach('gemini-3.8-flash', new MixGoogle({ options, config }));
     }
@@ -612,19 +587,9 @@ class ModelMix {
     grok46({ options = {}, config = {} } = {}) {
         return this.attach('grok-4.6', new MixGrok({ options, config }));
     }
-    grok45({ options = {}, config = {} } = {}) {
-        return this.attach('grok-4.5', new MixGrok({ options, config }));
-    }
     grok43({ options = {}, config = {} } = {}) {
         return this.attach('grok-4.3', new MixGrok({ options, config }));
     }
-    grok420multiAgent({ options = {}, config = {} } = {}) {
-        return this.attach('grok-4.20-multi-agent-0309', new MixGrok({ options, config }));
-    }
-    /** Non-reasoning by default; with `.effort(20+)` / `-1` resolves to the reasoning model at request time. */
-    grok420({ options = {}, config = {} } = {}) {
-        return this.attach('grok-4.20-0309', new MixGrok({ options, config }));
-    }
 
     museGlimmer30b({ options = {}, config = {}, mix = { fireworks: true } } = {}) {
         mix = { ...this.mix, ...mix };
@@ -726,14 +691,6 @@ class ModelMix {
         return this;
     }
 
-    kimiK25({ options = {}, config = {}, mix = { together: true } } = {}) {
-        mix = { ...this.mix, ...mix };
-        if (mix.together) this.attach('moonshotai/Kimi-K2.5', new MixTogether({ options, config }));
-        if (mix.fireworks) this.attach('accounts/fireworks/models/kimi-k2p5', new MixFireworks({ options, config }));
-        if (mix.openrouter) this.attach('moonshotai/kimi-k2.5', new MixOpenRouter({ options, config }));
-        return this;
-    }
-
     lmstudio(model = 'lmstudio', { options = {}, config = {} } = {}) {
         return this.attach(model, new MixLMStudio({ options, config }));
     }
@@ -1798,7 +1755,7 @@ class ModelMix {
             });
         } else {
             if (!this.models || this.models.length === 0) {
-                throw new Error('No models specified. Use methods like .gpt5(), .sonnet46() first.');
+                throw new Error('No models specified. Use methods like .gpt5mini(), .sonnet5() first.');
             }
             execution = this.limiter.schedule(() => {
                 throwIfAborted(signal);

package/lib/model-chain.js

@@ -1,19 +1,19 @@
 const { normalizeEffort } = require('../effort');
 
 const CHAIN_MODEL_SHORTCUTS = new Set([
-    'gpt5', 'gpt5mini', 'gpt5nano',
-    'gpt51', 'gpt52', 'gpt54', 'gpt54mini', 'gpt54nano', 'gpt54pro',
-    'gpt6astra', 'gpt55', 'gpt55pro', 'gpt56sol', 'gpt56terra', 'gpt56luna',
-    'gptRealtime', 'gptRealtimeMini', 'gpt53codex', 'gpt53chat', 'gptOss',
+    'gpt5mini', 'gpt5nano',
+    'gpt52', 'gpt54', 'gpt54mini', 'gpt54nano', 'gpt54pro',
+    'gpt6astra', 'gpt6sol', 'gpt6luna', 'gpt55', 'gpt55pro', 'gpt56sol', 'gpt56terra', 'gpt56luna',
+    'gptRealtime', 'gptRealtimeMini',
     'fable51', 'fable50', 'fable5', 'opus55', 'opus50', 'opus5', 'opus48', 'opus47', 'opus46',
-    'sonnet50', 'sonnet5', 'sonnet46', 'sonnet45', 'haiku45',
-    'gemini31pro', 'gemini38flash', 'gemini37flash', 'gemini36flash', 'gemini35flash',
+    'sonnet50', 'sonnet5', 'sonnet45', 'haiku45',
+    'gemini38flash', 'gemini37flash', 'gemini36flash', 'gemini35flash',
     'gemini35flashLite', 'gemini31flashLite', 'sonarPro', 'sonar',
-    'grok47', 'grok46', 'grok45', 'grok43', 'grok420multiAgent', 'grok420',
+    'grok47', 'grok46', 'grok43',
     'museGlimmer30b', 'museSpark12', 'museSpark12c', 'museSpark13', 'museSpark13c',
     'qwen35397b', 'qwen36plus', 'qwen37plus', 'qwen38max', 'qwen3827b', 'qwen38flash',
     'hermes470b', 'hermes4405b', 'hermes3',
-    'kimiK26', 'kimiK27Code', 'kimiK3', 'kimiK25',
+    'kimiK26', 'kimiK27Code', 'kimiK3',
     'minimaxM27', 'minimaxM3', 'mimo25', 'mimo25pro', 'mimo26pro',
     'deepseekV4Pro', 'deepseekPro', 'deepseekV4Flash', 'deepseekV41Flash', 'GLM52', 'GLM53', 'GLM53Flash'
 ]);

package/lib/providers/openai-options.js

@@ -3,7 +3,7 @@ function normalizeOpenAIOptions(options) {
         delete options.max_tokens;
         delete options.temperature;
     }
-    if (options.model?.includes('gpt-5') || /^(?:openai\/)?gpt-6-astra(?:-|$)/.test(options.model || '')) {
+    if (options.model?.includes('gpt-5') || /^(?:openai\/)?gpt-6(?:-|$)/.test(options.model || '')) {
         if (options.max_tokens) {
             options.max_completion_tokens = options.max_tokens;
             delete options.max_tokens;

package/lib/providers/openai.js

@@ -11,6 +11,12 @@ const {
     validateProviderExecution
 } = require('../abort-signal');
 
+/** Explicit prompt cache controls belong to the documented "GPT-5.6 and later" family. */
+function isGPT56OrLater(model) {
+    if (typeof model !== 'string') return false;
+    return model.startsWith('gpt-5.6') || /^(?:openai\/)?gpt-6(?:-|$)/.test(model);
+}
+
 function createOpenAIProviders({
     ModelMix,
     MixCustom,
@@ -38,9 +44,8 @@ function createOpenAIProviders({
         }
     
         static buildResponsesRequest(options = {}, config = {}) {
-            const isGPT56 = typeof options.model === 'string' && options.model.startsWith('gpt-5.6');
             const input = MixOpenAIResponses.messagesToResponsesInput(options.messages, {
-                translateNeutralCache: isGPT56
+                translateNeutralCache: isGPT56OrLater(options.model)
             });
             if (config.system) {
                 input.unshift({ role: 'developer', content: [{ type: 'input_text', text: config.system }] });
@@ -106,7 +111,7 @@ function createOpenAIProviders({
         }
     
         static validatePromptCaching(options, input) {
-            const isGPT56 = typeof options.model === 'string' && options.model.startsWith('gpt-5.6');
+            const supportsCacheControls = isGPT56OrLater(options.model);
             const cacheOptions = options.prompt_cache_options;
             const breakpoints = input.flatMap(message => Array.isArray(message.content)
                 ? message.content
@@ -114,14 +119,14 @@ function createOpenAIProviders({
                     .map(block => block.prompt_cache_breakpoint)
                 : []);
     
-            if (isGPT56 && options.prompt_cache_retention !== undefined) {
-                throw new Error('GPT-5.6 does not support prompt_cache_retention; use prompt_cache_options.ttl instead.');
+            if (supportsCacheControls && options.prompt_cache_retention !== undefined) {
+                throw new Error('GPT-5.6 and later models do not support prompt_cache_retention; use prompt_cache_options.ttl instead.');
             }
-            if (!isGPT56 && cacheOptions !== undefined) {
-                throw new Error('prompt_cache_options is only supported by GPT-5.6 models.');
+            if (!supportsCacheControls && cacheOptions !== undefined) {
+                throw new Error('prompt_cache_options is only supported by GPT-5.6 and later models.');
             }
-            if (!isGPT56 && breakpoints.length > 0) {
-                throw new Error('prompt_cache_breakpoint is only supported by GPT-5.6 models.');
+            if (!supportsCacheControls && breakpoints.length > 0) {
+                throw new Error('prompt_cache_breakpoint is only supported by GPT-5.6 and later models.');
             }
             if (cacheOptions !== undefined) {
                 if (!isPlainObject(cacheOptions)) {

package/lib/token-usage.js

@@ -22,9 +22,11 @@ function usesLongContextRates(pricing, inputTokens) {
 const MODEL_PRICING = {
     // OpenAI
     'gpt-6-astra': { input: 10.00, cachedInput: 1.00, cacheWrite: 12.50, output: 50.00, longContext: OPENAI_LONG_CONTEXT_PRICING },
+    'gpt-6-sol': { input: 2.00, cachedInput: 0.20, cacheWrite: 2.50, output: 10.00, longContext: OPENAI_LONG_CONTEXT_PRICING },
+    'gpt-6-luna': { input: 0.10, cachedInput: 0.01, cacheWrite: 0.125, output: 0.50, longContext: OPENAI_LONG_CONTEXT_PRICING },
     'gpt-realtime-mini': { input: 0.60, cachedInput: 0.06, output: 2.40 },
     'gpt-realtime': { input: 4.00, cachedInput: 0.40, output: 16.00 },
-    'gpt-5.6-sol': { input: 5.00, cachedInput: 0.50, cacheWrite: 6.25, output: 30.00, longContext: OPENAI_LONG_CONTEXT_PRICING },
+    'gpt-5.6-sol': { input: 4.00, cachedInput: 0.40, cacheWrite: 5.00, output: 20.00, longContext: OPENAI_LONG_CONTEXT_PRICING },
     'gpt-5.6-terra': { input: 2.00, cachedInput: 0.20, cacheWrite: 2.50, output: 12.00, longContext: OPENAI_LONG_CONTEXT_PRICING },
     'gpt-5.6-luna': { input: 0.20, cachedInput: 0.02, cacheWrite: 0.25, output: 1.20, longContext: OPENAI_LONG_CONTEXT_PRICING },
     'gpt-5.5-pro': { input: 30.00, output: 180.00 },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "modelmix",
-  "version": "5.2.3",
+  "version": "5.3.1",
   "description": "🧬 Reliable interface with automatic fallback for AI LLMs.",
   "main": "index.js",
   "types": "index.d.ts",