modelmix 5.1.10 → 5.1.15

package/README.md

@@ -95,6 +95,26 @@ const model = await ModelMix.new(setup)
 console.log(await model.message());
 ```
 
+**Cancel one execution with AbortSignal**
+
+Pass the signal directly to the terminal method. ModelMix propagates it through
+queued work, retries, fallbacks, provider transports, plugins, child executions,
+and tool calls:
+
+```javascript
+const controller = new AbortController();
+const response = model.message(controller.signal);
+
+controller.abort(new Error('Request cancelled'));
+await response;
+```
+
+Use `raw(signal)`, `stream(callback, signal)`, `block(options, signal)`, or
+`json(example, description, jsonOptions, signal)` for the other output modes.
+`execute()` accepts it as the top-level `signal` field. Never put a signal in
+ModelMix or provider `config`/`options`; those locations are rejected because
+they are reusable and may become provider payload fields.
+
 The same ordered chain can be attached by passing model shortcuts directly to
 `chain()`. Add `@effort` to override unified effort for one model; entries
 without it inherit the chain effort, or use the provider default when the chain
@@ -119,10 +139,9 @@ 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.**
 ```javascript
-ModelMix.new({ mix: { openrouter: false } })
+ModelMix.new()
   .gptOss()
   .kimiK25()
-  .hermes3()
   .addText('What is the capital of France?');
 ```
 
@@ -143,15 +162,20 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `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] |
+| `gpt55pro()` | OpenAI | gpt-5.5-pro | [\$30.00][1] | [\$180.00][1] |
 | `gpt54()` | OpenAI | gpt-5.4 | [\$2.50][1] | [\$15.00][1] |
 | `gpt54mini()` | OpenAI | gpt-5.4-mini | [\$0.75][1] | [\$4.50][1] |
 | `gpt54nano()` | OpenAI | gpt-5.4-nano | [\$0.20][1] | [\$1.25][1] |
-| `gpt53codex()` | OpenAI | gpt-5.3-codex | [\$1.25][1] | [\$14.00][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] |
 | `opus5()` | Anthropic | claude-opus-5 | [\$5.00][2] | [\$25.00][2] |
 | `opus48()` | Anthropic | claude-opus-4-8 | [\$5.00][2] | [\$25.00][2] |
@@ -172,6 +196,7 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `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] |
+| `museSpark12Contributor()` | OpenRouter | meta/muse-spark-1.2-contributor | [\$0.10][22] | [\$0.20][22] |
 | `qwen35397b()` | OpenRouter | qwen/qwen3.5-397b-a17b | [\$0.385][14] | [\$2.45][14] |
 | `qwen36plus()` | OpenRouter | qwen/qwen3.6-plus | [\$0.325][18] | [\$1.95][18] |
 | `qwen37plus()` | Fireworks | models/qwen3p7-plus | [\$0.40][10] | [\$1.60][10] |
@@ -197,9 +222,13 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 
 Gemini 3.7 Flash and 3.6 Flash use Google's introductory standard pricing through December 31, 2026; standard rates double on January 1, 2027.
 
-`museGlimmer30b()` uses Fireworks first and OpenRouter as its default fallback. NVIDIA NIM and Together are also available through `mix.nvidia` and `mix.together`; disable either default provider with `mix.fireworks: false` or `mix.openrouter: false`.
+`museGlimmer30b()` uses Fireworks by default. OpenRouter, NVIDIA NIM, and Together are available through `mix.openrouter`, `mix.nvidia`, and `mix.together`.
+
+`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 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.
 
-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 enabled provider in its fallback chain. `qwen36plus()` now defaults to OpenRouter because Fireworks retired its serverless deployment; Fireworks remains available explicitly for private or on-demand deployments.
+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.
 
 [1]: https://platform.openai.com/docs/pricing "Pricing | OpenAI"
 [2]: https://docs.anthropic.com/en/docs/about-claude/pricing "Pricing - Anthropic"
@@ -221,6 +250,9 @@ The multi-provider shortcuts also expose the current catalog alternatives: `gptO
 [18]: https://openrouter.ai/qwen/qwen3.6-plus "Qwen 3.6 Plus on OpenRouter"
 [19]: https://openrouter.ai/qwen/qwen3.8-flash "Qwen3.8 Flash on OpenRouter"
 [20]: https://openrouter.ai/z-ai/glm-5.3-flash "GLM 5.3 Flash on OpenRouter"
+[21]: https://openrouter.ai/anthropic/claude-fable-5.1 "Claude Fable 5.1 on OpenRouter"
+[22]: https://openrouter.ai/meta/muse-spark-1.2-contributor "Muse Spark 1.2 Contributor on OpenRouter"
+[23]: https://openrouter.ai/openai "OpenAI Models on OpenRouter"
 
 Each method accepts optional `options`, `config`, and (for multi-provider methods) `mix` parameters to customize behavior.  
 
@@ -927,7 +959,7 @@ const model = ModelMix.new()
     .addText('Summarize this request.');
 ```
 
-A plugin may edit `context.request`, call `next()`, or return a complete ModelMix result itself. It can also create history-free child executions with `context.invoke()` and choose plugin inheritance:
+A plugin may edit `context.request`, call `next()`, or return a complete ModelMix result itself. The current execution signal is available as `context.signal`, and child executions created with `context.invoke()` inherit it automatically. Local tool callbacks receive the same signal as their second argument. A plugin can also choose plugin inheritance:
 
 ```javascript
 const child = await context.invoke({
@@ -1038,8 +1070,8 @@ new ModelMix(args = { options: {}, config: {} })
 - `assign(keyValues)`: Assigns EJS data for messages and system prompts.
 - `assignKey(key, value)`: Assigns one EJS data value.
 - `assignKeyFromFile(key, filePath)`: Renders an EJS file through `include` and assigns its output to one key.
-- `message()`: Sends the message and returns the response.
-- `raw()`: Sends the message and returns the complete response data including:
+- `message(signal?)`: Sends the message and returns the response.
+- `raw(signal?)`: Sends the message and returns the complete response data including:
   - `message`: The text response from the model
   - `think`: Reasoning/thinking content (if available)
   - `toolCalls`: Array of tool calls made by the model (if any)
@@ -1057,8 +1089,8 @@ new ModelMix(args = { options: {}, config: {} })
 
   if (profile.flagged) throw new Error('Profile rejected by moderation');
   ```
-- `stream(callback)`: Sends the message and streams the response, invoking the callback with each streamed part.
-- `json(schemaExample, descriptions = {}, options = {})`: Forces the model to return a response in a specific JSON format.
+- `stream(callback, signal?)`: Sends the message and streams the response, invoking the callback with each streamed part.
+- `json(schemaExample, descriptions = {}, options = {}, signal?)`: Forces the model to return a response in a specific JSON format.
   - `schemaExample`: Example of the JSON structure to be returned. Top-level arrays are auto-wrapped for better LLM compatibility.
   - `descriptions`: Descriptions for each field — can be strings or descriptor objects with `{ description, required, enum, default }`.
   - `options`: `{ addSchema: true, addExample: false, addNote: false }`

package/demo/free.js

@@ -1,12 +1,10 @@
 import { ModelMix } from '../index.js';
 try { process.loadEnvFile(); } catch {}
 
-const ai = ModelMix.new({ config: { debug: 2 }, mix: { openrouter: false } })
+const ai = ModelMix.new({ config: { debug: 2 } })
     .gptOss()
     .kimiK25()
-    .hermes3()
     .addText('What is the capital of France?');
 
 const response = await ai.message();
 console.log('Response from a free-tier provider:', response);
-

package/demo/short.js

@@ -9,19 +9,21 @@ const setup = {
 };
 
 const mmix = await ModelMix.new(setup)
-    .sonnet46() // (main model) Anthropic claude-sonnet-4-6
-    .gpt56luna() // (fallback 1) OpenAI gpt-5.6-luna
-    .gemini37flash() // (fallback 2) Google gemini-3.7-flash
-    .gpt5nano() // (fallback 3) OpenAI gpt-5-nano
-    .grok46() // (fallback 4) Grok grok-4.6
-    .qwen35397b() // (fallback 5) OpenRouter qwen/qwen3.5-397b-a17b
-    .qwen3827b() // (fallback 6) OpenRouter qwen/qwen3.8-27b
-    .qwen38flash() // (fallback 7) OpenRouter qwen/qwen3.8-flash
-    .GLM53() // (fallback 8) OpenRouter z-ai/glm-5.3
-    .GLM53Flash() // (fallback 9) OpenRouter z-ai/glm-5.3-flash
-    .museGlimmer30b({ mix: { fireworks: false, openrouter: true } }) // (fallback 10) OpenRouter meta/muse-glimmer-30b
-    .hermes470b() // (fallback 11) OpenRouter nousresearch/hermes-4-70b
-    .hermes4405b() // (fallback 12) OpenRouter nousresearch/hermes-4-405b
+    .fable51({ mix: { openrouter: true } }) // (main + provider fallback) Anthropic/OpenRouter Claude Fable 5.1
+    .sonnet46() // (fallback 2) Anthropic claude-sonnet-4-6
+    .gpt56luna({ mix: { openrouter: true } }) // (fallback 3 + provider fallback) OpenAI/OpenRouter gpt-5.6-luna
+    .gemini37flash() // (fallback 4) Google gemini-3.7-flash
+    .gpt5nano({ mix: { openrouter: true } }) // (fallback 5 + provider fallback) OpenAI/OpenRouter gpt-5-nano
+    .grok46() // (fallback 6) Grok grok-4.6
+    .qwen35397b() // (fallback 7) OpenRouter qwen/qwen3.5-397b-a17b
+    .qwen3827b() // (fallback 8) OpenRouter qwen/qwen3.8-27b
+    .qwen38flash() // (fallback 9) OpenRouter qwen/qwen3.8-flash
+    .GLM53() // (fallback 10) OpenRouter z-ai/glm-5.3
+    .GLM53Flash() // (fallback 11) OpenRouter z-ai/glm-5.3-flash
+    .museGlimmer30b({ mix: { fireworks: false, openrouter: true } }) // (fallback 12) OpenRouter meta/muse-glimmer-30b
+    .museSpark12Contributor() // (fallback 13) OpenRouter meta/muse-spark-1.2-contributor
+    .hermes470b() // (fallback 14) OpenRouter nousresearch/hermes-4-70b
+    .hermes4405b() // (fallback 15) OpenRouter nousresearch/hermes-4-405b
     .addText("What's your name?");
 
 console.log(await mmix.message());

package/effort.js

@@ -7,7 +7,7 @@
  */
 
 const OPENAI_LEVELS = ['none', 'low', 'medium', 'high', 'xhigh'];
-const OPENAI_LEVEL_LADDER = [...OPENAI_LEVELS, 'max'];
+const OPENAI_LEVEL_LADDER = ['none', 'minimal', 'low', 'medium', 'high', 'xhigh', 'max'];
 const ANTHROPIC_LEVELS = ['low', 'medium', 'high', 'xhigh', 'max'];
 const GEMINI_LEVELS = ['minimal', 'low', 'medium', 'high'];
 
@@ -45,9 +45,11 @@ const OPENAI_MODEL_LEVELS = {
     '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'],
+    'anthropic/claude-fable-5.1': ['low', 'medium', 'high', 'xhigh', 'max'],
     'meta/muse-glimmer-30b': ['low', 'medium', 'high'],
     'accounts/fireworks/models/muse-glimmer-30b': ['low', 'medium', 'high', 'xhigh'],
     'meta-models/Muse-Glimmer-30B': ['low', 'medium', 'high', 'xhigh'],
+    'meta/muse-spark-1.2-contributor': ['minimal', 'low', 'medium', 'high', 'xhigh'],
     'accounts/fireworks/models/qwen3p8-2p4t-a95b': ['none', 'low', 'medium', 'high'],
     'qwen/qwen3.8-27b': ['low', 'medium', 'xhigh'],
     'qwen/qwen3.8-flash': ['low', 'medium', 'xhigh'],
@@ -87,7 +89,6 @@ const DEEPSEEK_BANDS = [
  * MiniMax M3 (OpenAI-compatible): thinking.type disabled | adaptive.
  * @see https://platform.minimax.io/docs/api-reference/text-openai-api
  */
-const MINIMAX_LEVELS = ['disabled', 'adaptive'];
 const MINIMAX_BANDS = [
     [0, 19, 'disabled'],
     [20, 100, 'adaptive'],
@@ -203,8 +204,11 @@ function pickNearestLevel(desired, ladder, supported) {
 }
 
 function supportedOpenAILevels(modelKey) {
-    if (modelKey && OPENAI_MODEL_LEVELS[modelKey]) {
-        return OPENAI_MODEL_LEVELS[modelKey];
+    const canonicalModelKey = typeof modelKey === 'string' && modelKey.startsWith('openai/')
+        ? modelKey.slice('openai/'.length)
+        : modelKey;
+    if (canonicalModelKey && OPENAI_MODEL_LEVELS[canonicalModelKey]) {
+        return OPENAI_MODEL_LEVELS[canonicalModelKey];
     }
     return OPENAI_LEVELS;
 }
@@ -400,9 +404,11 @@ function mapEffort(providerFamily, effort, modelKey) {
         const supported = supportedOpenAILevels(modelKey);
         const desired = modelKey === 'z-ai/glm-5.3' || modelKey === 'z-ai/glm-5.3-flash'
             ? levelFromBands(normalized, GLM53_BANDS)
-            : normalized === 100 && supported.includes('max')
-                ? 'max'
-                : levelFromBands(normalized, OPENAI_BANDS);
+            : modelKey === 'anthropic/claude-fable-5.1'
+                ? levelFromBands(normalized, ANTHROPIC_BANDS)
+                : normalized === 100 && supported.includes('max')
+                    ? 'max'
+                    : levelFromBands(normalized, OPENAI_BANDS);
         const level = pickNearestLevel(desired, OPENAI_LEVEL_LADDER, supported);
         return { reasoning_effort: level };
     }

package/http-client.js

@@ -45,8 +45,8 @@ async function buildHttpError(url, response) {
     return error;
 }
 
-async function fetchJsonResponse(url, { method = 'POST', headers = {}, body } = {}) {
-    const response = await fetch(url, { method, headers, body });
+async function fetchJsonResponse(url, { method = 'POST', headers = {}, body, signal } = {}) {
+    const response = await fetch(url, { method, headers, body, signal });
     if (!response.ok) {
         throw await buildHttpError(url, response);
     }
@@ -58,8 +58,8 @@ async function fetchJsonResponse(url, { method = 'POST', headers = {}, body } =
     };
 }
 
-async function fetchBinaryResponse(url, { method = 'GET', headers = {}, body } = {}) {
-    const response = await fetch(url, { method, headers, body });
+async function fetchBinaryResponse(url, { method = 'GET', headers = {}, body, signal } = {}) {
+    const response = await fetch(url, { method, headers, body, signal });
     if (!response.ok) {
         throw await buildHttpError(url, response);
     }
@@ -71,8 +71,8 @@ async function fetchBinaryResponse(url, { method = 'GET', headers = {}, body } =
     };
 }
 
-async function fetchStreamResponse(url, { method = 'POST', headers = {}, body } = {}) {
-    const response = await fetch(url, { method, headers, body });
+async function fetchStreamResponse(url, { method = 'POST', headers = {}, body, signal } = {}) {
+    const response = await fetch(url, { method, headers, body, signal });
     if (!response.ok) {
         throw await buildHttpError(url, response);
     }

package/index.d.ts

@@ -76,6 +76,7 @@ export interface ModelMixConfig {
 }
 
 export interface ModelMixMixFlags {
+  anthropic?: boolean;
   openrouter?: boolean;
   cerebras?: boolean;
   groq?: boolean;
@@ -258,6 +259,7 @@ export interface PluginExecutionContext {
     outputMode: ModelMixOutputMode;
   };
   execution: Readonly<PluginExecutionMetadata>;
+  signal?: AbortSignal;
   invoke(input: ChildInvocation): Promise<ModelMixResult>;
 }
 
@@ -311,7 +313,8 @@ export interface ToolDefinition {
 }
 
 export type ToolCallback = (
-  args: Record<string, unknown>
+  args: Record<string, unknown>,
+  signal?: AbortSignal
 ) => unknown | Promise<unknown>;
 
 export interface ToolWithCallback {
@@ -342,6 +345,7 @@ export interface ProviderConstructorArgs {
 export interface CreateArgs {
   config?: ModelMixConfig;
   options?: ModelMixOptions;
+  signal?: AbortSignal;
   outputMode?: ModelMixOutputMode;
 }
 
@@ -462,6 +466,7 @@ export declare class ModelMix {
   gptOss(args?: ModelAttachArgs): this;
 
   // Anthropic
+  fable51(args?: ModelAttachArgs): this;
   fable50(args?: ModelAttachArgs): this;
   fable5(args?: ModelAttachArgs): this;
   opus50(args?: ModelAttachArgs): this;
@@ -497,6 +502,7 @@ export declare class ModelMix {
 
   // Multi-provider
   museGlimmer30b(args?: ModelAttachArgs): this;
+  museSpark12Contributor(args?: ModelAttachArgs): this;
   qwen35397b(args?: ModelAttachArgs): this;
   qwen36plus(args?: ModelAttachArgs): this;
   qwen37plus(args?: ModelAttachArgs): this;
@@ -530,15 +536,16 @@ export declare class ModelMix {
   addImageFromUrl(url: string, options?: RoleOptions): Promise<this>;
   processImages(): Promise<void>;
 
-  message(): Promise<string>;
+  message(signal?: AbortSignal): Promise<string>;
   json<T = unknown>(
     schemaExample?: T | T[] | null,
     schemaDescription?: SchemaDescription,
-    options?: JsonMethodOptions
+    options?: JsonMethodOptions,
+    signal?: AbortSignal
   ): Promise<T>;
-  block(options?: BlockOptions): Promise<string>;
-  raw(): Promise<ModelMixResult>;
-  stream(callback: StreamCallback): Promise<ModelMixResult>;
+  block(options?: BlockOptions, signal?: AbortSignal): Promise<string>;
+  raw(signal?: AbortSignal): Promise<ModelMixResult>;
+  stream(callback: StreamCallback, signal?: AbortSignal): Promise<ModelMixResult>;
 
   assignKeyFromFile(key: string, filePath: string): this;
   groupByRoles(messages: ChatMessage[]): ChatMessage[];