modelmix 5.1.19 → 5.1.20

package/README.md

@@ -158,6 +158,7 @@ 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] |
 | `gpt56terra()` | OpenAI | gpt-5.6-terra | [\$2.00][1] | [\$12.00][1] |
 | `gpt56luna()` | OpenAI | gpt-5.6-luna | [\$0.20][1] | [\$1.20][1] |
@@ -230,7 +231,7 @@ 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 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. `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.
 
 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.
 
@@ -302,6 +303,7 @@ ModelMix.new().effort(-1).minimaxM3().addText('...').message();
 ### Provider-specific behavior
 
 - **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-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.

package/demo/short.js

@@ -27,6 +27,7 @@ const mmix = await ModelMix.new(setup)
     .museSpark13() // (fallback 16) OpenRouter meta/muse-spark-1.3
     .museSpark12() // (fallback 17) OpenRouter meta/muse-spark-1.2
     .museSpark13c() // (fallback 18) OpenRouter meta/muse-spark-1.3-contributor
+    .gpt6astra() // (fallback 19) OpenAI gpt-6-astra
     .addText("What's your name?");
 
 console.log(await mmix.message());

package/effort.js

@@ -42,6 +42,7 @@ const GLM53_BANDS = [
 
 /** Exact model → supported OpenAI reasoning_effort values */
 const OPENAI_MODEL_LEVELS = {
+    'gpt-6-astra': ['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'],

package/index.d.ts

@@ -456,6 +456,7 @@ export declare class ModelMix {
   gpt54pro(args?: ModelAttachArgs): this;
   gpt55(args?: ModelAttachArgs): this;
   gpt55pro(args?: ModelAttachArgs): this;
+  gpt6astra(args?: ModelAttachArgs): this;
   gpt56sol(args?: ModelAttachArgs): this;
   gpt56terra(args?: ModelAttachArgs): this;
   gpt56luna(args?: ModelAttachArgs): this;

package/index.js

@@ -494,6 +494,9 @@ class ModelMix {
     gpt55pro(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-5.5-pro', MixOpenAIResponses, args);
     }
+    gpt6astra(args = {}) {
+        return this._attachOpenAIWithOpenRouter('gpt-6-astra', MixOpenAIResponses, args);
+    }
     gpt56sol(args = {}) {
         return this._attachOpenAIWithOpenRouter('gpt-5.6-sol', MixOpenAIResponses, args);
     }

package/lib/model-chain.js

@@ -3,7 +3,7 @@ const { normalizeEffort } = require('../effort');
 const CHAIN_MODEL_SHORTCUTS = new Set([
     'gpt5', 'gpt5mini', 'gpt5nano',
     'gpt51', 'gpt52', 'gpt54', 'gpt54mini', 'gpt54nano', 'gpt54pro',
-    'gpt55', 'gpt55pro', 'gpt56sol', 'gpt56terra', 'gpt56luna',
+    'gpt6astra', 'gpt55', 'gpt55pro', 'gpt56sol', 'gpt56terra', 'gpt56luna',
     'gptRealtime', 'gptRealtimeMini', 'gpt53codex', 'gpt53chat', 'gptOss',
     'fable51', 'fable50', 'fable5', 'opus50', 'opus5', 'opus48', 'opus47', 'opus46',
     'sonnet50', 'sonnet5', 'sonnet46', 'sonnet45', 'haiku45',

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')) {
+    if (options.model?.includes('gpt-5') || /^(?:openai\/)?gpt-6-astra(?:-|$)/.test(options.model || '')) {
         if (options.max_tokens) {
             options.max_completion_tokens = options.max_tokens;
             delete options.max_tokens;

package/lib/token-usage.js

@@ -1,4 +1,4 @@
-const GPT56_LONG_CONTEXT_PRICING = Object.freeze({
+const OPENAI_LONG_CONTEXT_PRICING = Object.freeze({
     inputThreshold: 272_000,
     inputMultiplier: 2,
     outputMultiplier: 1.5
@@ -21,11 +21,12 @@ 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-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: GPT56_LONG_CONTEXT_PRICING },
-    'gpt-5.6-terra': { input: 2.00, cachedInput: 0.20, cacheWrite: 2.50, output: 12.00, longContext: GPT56_LONG_CONTEXT_PRICING },
-    'gpt-5.6-luna': { input: 0.20, cachedInput: 0.02, cacheWrite: 0.25, output: 1.20, longContext: GPT56_LONG_CONTEXT_PRICING },
+    'gpt-5.6-sol': { input: 5.00, cachedInput: 0.50, cacheWrite: 6.25, output: 30.00, 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 },
     'gpt-5.5': { input: 5.00, cachedInput: 0.50, output: 30.00 },
     'gpt-5.4': { input: 2.50, cachedInput: 0.25, output: 15.00 },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "modelmix",
-  "version": "5.1.19",
+  "version": "5.1.20",
   "description": "🧬 Reliable interface with automatic fallback for AI LLMs.",
   "main": "index.js",
   "types": "index.d.ts",
@@ -79,6 +79,5 @@
     "test:plugins": "mocha test/plugins.test.js --timeout 10000 --require test/setup.js",
     "test:rlm": "mocha plugins/rlm/test/**/*.test.js --timeout 10000 --require test/setup.js",
     "test:offline": "mocha test/abort.test.js test/json.test.js test/fallback.test.js test/templates.test.js test/images.test.js test/bottleneck.test.js test/tokens.test.js test/history.test.js test/anthropic.test.js test/effort.test.js test/grok.test.js test/moderation.test.js test/plugins.test.js plugins/rlm/test/**/*.test.js --timeout 10000 --require test/setup.js"
-  },
-  "packageManager": "pnpm@11.18.0+sha512.33d83c77da82f49fba836925c6f1b841181ec3132b670639bd012f7075f5c7cf634c5f870147c19aae7478fac01df09d8892e880454896edd23ee9b33757563c"
-}
+  }
+}

package/skills/modelmix/SKILL.md

@@ -143,7 +143,7 @@ ModelMix.new({ config: { effort: 80 } })
 | DeepSeek V4 | off | `low`↑ | `high`↑ | `high`↑ | `max`↑ | — |
 | MiniMax M3 | off | adaptive | adaptive | adaptive | adaptive | adaptive |
 
-\* GPT-5.6 maps `100` to `max`; 80–99 remains `xhigh`. Qwen 3.8 27B and Flash map 0–39 / 40–79 / 80–100 to `low` / `medium` / `xhigh`; Qwen 3.8 Flash is the managed production version based on Flash-Next. GLM 5.3 and GLM 5.3 Flash require reasoning and map those bands to `low` / `high` / `max`. Gemini bands: 0–24 / 25–49 / 50–74 / 75–100. Gemini 3.8 Flash and 3.7 Flash support only `low` / `medium` / `high`, so the first two bands clamp to `low`; `-1` keeps their native `medium` default. DeepSeek `↑` = thinking on; `off` = thinking disabled. MiniMax `off`/`adaptive` = `thinking.disabled` / `thinking.type=adaptive`. Gemini 2.5 maps 0–100 to `thinkingBudget`. Anthropic: adaptive + `output_config.effort` on Claude 5 / Fable / Opus 4.6+ / Sonnet 4.6+; Sonnet 4.5 / Haiku 4.5 use `thinking.type=enabled` + `budget_tokens`. Grok 4.6 maps 0–39 / 40–59 / 60–79 / 80–100 to `low` / `medium` / `high` / `xhigh`; without effort it uses native `high`. `-1` = adaptive/dynamic when available, else no-op. Levels clamp per model. Former `*think()` methods are removed — use `.effort(n).<model>()`. Kimi: `kimiK25()` / `kimiK26()`. Grok 4.20: `.grok420()` non-reasoning; `.effort(20+|-1).grok420()` selects reasoning.
+\* GPT-6 Astra maps 0–39 / 40–59 / 60–79 / 80–99 / 100 to `low` / `medium` / `high` / `xhigh` / `max`. GPT-5.6 maps `100` to `max`; 80–99 remains `xhigh`. Qwen 3.8 27B and Flash map 0–39 / 40–79 / 80–100 to `low` / `medium` / `xhigh`; Qwen 3.8 Flash is the managed production version based on Flash-Next. GLM 5.3 and GLM 5.3 Flash require reasoning and map those bands to `low` / `high` / `max`. Gemini bands: 0–24 / 25–49 / 50–74 / 75–100. Gemini 3.8 Flash and 3.7 Flash support only `low` / `medium` / `high`, so the first two bands clamp to `low`; `-1` keeps their native `medium` default. DeepSeek `↑` = thinking on; `off` = thinking disabled. MiniMax `off`/`adaptive` = `thinking.disabled` / `thinking.type=adaptive`. Gemini 2.5 maps 0–100 to `thinkingBudget`. Anthropic: adaptive + `output_config.effort` on Claude 5 / Fable / Opus 4.6+ / Sonnet 4.6+; Sonnet 4.5 / Haiku 4.5 use `thinking.type=enabled` + `budget_tokens`. Grok 4.6 maps 0–39 / 40–59 / 60–79 / 80–100 to `low` / `medium` / `high` / `xhigh`; without effort it uses native `high`. `-1` = adaptive/dynamic when available, else no-op. Levels clamp per model. Former `*think()` methods are removed — use `.effort(n).<model>()`. Kimi: `kimiK25()` / `kimiK26()`. Grok 4.20: `.grok420()` non-reasoning; `.effort(20+|-1).grok420()` selects reasoning.
 
 ## Available Model Shorthands
 
@@ -151,9 +151,9 @@ ModelMix.new({ config: { effort: 80 } })
 
 Use `ModerationMix.new().openai()` with `.raw()` to classify text and images through OpenAI's Moderations endpoint. Read the results from `raw.moderation`. `ModerationMix` accepts moderation providers as ordered fallbacks, rejects generative providers, and does not generate text or support streaming.
 
-`gpt56sol()` `gpt56terra()` `gpt56luna()` `gpt55()` `gpt55pro()` `gpt54()` `gpt54mini()` `gpt54nano()` `gpt54pro()` `gpt53codex()` `gpt53chat()` `gpt52()` `gpt51()` `gpt5()` `gpt5mini()` `gpt5nano()` `gptRealtime()` `gptRealtimeMini()` `gptOss()`
+`gpt6astra()` `gpt56sol()` `gpt56terra()` `gpt56luna()` `gpt55()` `gpt55pro()` `gpt54()` `gpt54mini()` `gpt54nano()` `gpt54pro()` `gpt53codex()` `gpt53chat()` `gpt52()` `gpt51()` `gpt5()` `gpt5mini()` `gpt5nano()` `gptRealtime()` `gptRealtimeMini()` `gptOss()`
 
-Every textual GPT-5 shortcut registers only the official OpenAI model by default. Pass `mix: { openrouter: true }` to `ModelMix.new()` or to an individual shortcut to append its `openai/*` OpenRouter route as a fallback. `gpt53chat()` uses `gpt-5.3-chat-latest` officially and `openai/gpt-5.3-chat` through OpenRouter. Both API keys are required when that fallback is enabled. Realtime shortcuts remain official-only.
+Every textual GPT-5 and GPT-6 shortcut registers only the official OpenAI model by default. Pass `mix: { openrouter: true }` to `ModelMix.new()` or to an individual shortcut to append its `openai/*` OpenRouter route as a fallback. `gpt53chat()` uses `gpt-5.3-chat-latest` officially and `openai/gpt-5.3-chat` through OpenRouter. Both API keys are required when that fallback is enabled. Realtime shortcuts remain official-only.
 
 ### Anthropic
 `fable51()` `fable50()` `opus50()` `opus48()` `opus47()` `opus46()` `sonnet5()` `sonnet46()` `sonnet45()` `haiku45()`

package/test/effort.test.js

@@ -73,6 +73,15 @@ describe('Unified effort scale', () => {
             expect(mapEffort('openai', 100)).to.deep.equal({ reasoning_effort: 'xhigh' });
         });
 
+        it('clamps GPT-6 Astra effort to its supported range', () => {
+            for (const key of ['gpt-6-astra', 'openai/gpt-6-astra']) {
+                for (const [effort, level] of [[0, 'low'], [39, 'low'], [40, 'medium'], [60, 'high'], [80, 'xhigh'], [99, 'xhigh'], [100, 'max']]) {
+                    expect(mapEffort('openai', effort, key)).to.deep.equal({ reasoning_effort: level });
+                }
+                expect(mapEffort('openai', -1, key)).to.equal(null);
+            }
+        });
+
         it('maps GPT-5.6 maximum unified effort to max', () => {
             expect(mapEffort('openai', 99, 'gpt-5.6-luna')).to.deep.equal({ reasoning_effort: 'xhigh' });
             for (const model of ['gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna']) {

package/test/fallback.test.js

@@ -57,6 +57,7 @@ describe('Provider Fallback Chain Tests', () => {
         it('should register every supported OpenAI text shortcut with its OpenRouter fallback', () => {
             expect(ModelMix.new().mix.openrouter).to.equal(false);
             const shortcuts = [
+                ['gpt6astra', 'gpt-6-astra', 'openai/gpt-6-astra', MixOpenAIResponses],
                 ['gpt5', 'gpt-5', 'openai/gpt-5', MixOpenAI],
                 ['gpt5mini', 'gpt-5-mini', 'openai/gpt-5-mini', MixOpenAI],
                 ['gpt5nano', 'gpt-5-nano', 'openai/gpt-5-nano', MixOpenAI],
@@ -101,13 +102,13 @@ describe('Provider Fallback Chain Tests', () => {
         });
 
         it('should append OpenRouter GPT fallbacks in chain() only when enabled globally', () => {
-            const official = ModelMix.new().chain('gpt56sol@100');
-            expect(official.models.map(({ key }) => key)).to.deep.equal(['gpt-5.6-sol']);
+            const official = ModelMix.new().chain('gpt6astra@100');
+            expect(official.models.map(({ key }) => key)).to.deep.equal(['gpt-6-astra']);
 
-            const routed = ModelMix.new({ mix: { openrouter: true } }).chain('gpt56sol@100');
+            const routed = ModelMix.new({ mix: { openrouter: true } }).chain('gpt6astra@100');
             expect(routed.models.map(({ key }) => key)).to.deep.equal([
-                'gpt-5.6-sol',
-                'openai/gpt-5.6-sol'
+                'gpt-6-astra',
+                'openai/gpt-6-astra'
             ]);
             expect(routed.models.map(({ provider }) => provider.config.effort)).to.deep.equal([100, 100]);
         });
@@ -140,6 +141,40 @@ describe('Provider Fallback Chain Tests', () => {
             expect(openRouterRequest).to.not.have.property('temperature');
         });
 
+        it('should fallback from the official GPT-6 Astra endpoint to OpenRouter', async () => {
+            let openRouterRequest;
+            model.effort(100).gpt6astra({ mix: { openrouter: true } }).addText('Hello');
+
+            nock('https://api.openai.com')
+                .post('/v1/responses', body => {
+                    expect(body.model).to.equal('gpt-6-astra');
+                    expect(body.reasoning).to.deep.equal({ effort: 'max' });
+                    expect(body.max_output_tokens).to.equal(8192);
+                    expect(body).to.not.have.property('temperature');
+                    return true;
+                })
+                .reply(503, { error: 'Service unavailable' });
+
+            nock('https://openrouter.ai')
+                .post('/api/v1/chat/completions', body => {
+                    openRouterRequest = body;
+                    return true;
+                })
+                .reply(200, {
+                    choices: [{
+                        message: {
+                            role: 'assistant',
+                            content: 'Hello from GPT-6 Astra through OpenRouter!'
+                        }
+                    }]
+                });
+
+            expect(await model.message()).to.equal('Hello from GPT-6 Astra through OpenRouter!');
+            expect(openRouterRequest.model).to.equal('openai/gpt-6-astra');
+            expect(openRouterRequest.max_completion_tokens).to.equal(8192);
+            expect(openRouterRequest).to.not.have.property('temperature');
+        });
+
         it('should keep the default fable51 chain on Anthropic', () => {
             model.chain('fable51@80');
 

package/test/tokens.test.js

@@ -337,6 +337,14 @@ describe('Token Usage Tracking', () => {
         expect(model.models[3].provider).to.be.instanceOf(MixOpenRouter);
     });
 
+    it('should account for Astra cache usage and the long-context boundary', function () {
+        for (const key of ['gpt-6-astra', 'openai/gpt-6-astra']) {
+            const tokens = { input: 272_000, cached: 100_000, cacheWrite: 20_000, output: 1_000 };
+            expect(ModelMix.calculateCost(key, tokens)).to.be.closeTo(1.92, 1e-10);
+            expect(ModelMix.calculateCost(key, { ...tokens, input: 272_001 })).to.be.closeTo(3.81502, 1e-10);
+        }
+    });
+
     it('should register GPT-5.6 shortcuts with OpenAI Responses provider', function () {
         const model = ModelMix.new({ mix: { openrouter: true } })
             .gpt56sol()