modelmix 5.1.8 → 5.1.9

package/README.md

@@ -177,9 +177,11 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `qwen37plus()` | Fireworks | models/qwen3p7-plus | [\$0.40][10] | [\$1.60][10] |
 | `qwen38max()` | Fireworks | qwen3p8-2p4t-a95b | [\$2.00][10] | [\$6.00][10] |
 | `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] |
 | `deepseekV4Pro()` | Fireworks | models/deepseek-v4-pro-0813 | [\$1.32][12] | [\$3.96][12] |
 | `GLM53()` | OpenRouter | z-ai/glm-5.3 | [\$1.40][16] | [\$4.40][16] |
+| `GLM53Flash()` | OpenRouter | z-ai/glm-5.3-flash | [\$0.075][20] | [\$0.25][20] |
 | `GLM52()` | Together | zai-org/GLM-5.2 | [\$1.40][7] | [\$4.40][7] |
 | `minimaxM3()` | MiniMax | MiniMax-M3 | [\$0.30][9] | [\$1.20][9] |
 | `minimaxM27()` | MiniMax | MiniMax-M2.7 | [\$0.30][9] | [\$1.20][9] |
@@ -217,6 +219,8 @@ The multi-provider shortcuts also expose the current catalog alternatives: `gptO
 [16]: https://openrouter.ai/z-ai/glm-5.3 "GLM 5.3 on OpenRouter"
 [17]: https://fireworks.ai/models/fireworks/muse-glimmer-30b "Muse Glimmer 30B on Fireworks"
 [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"
 
 Each method accepts optional `options`, `config`, and (for multi-provider methods) `mix` parameters to customize behavior.  
 
@@ -257,8 +261,8 @@ ModelMix.new().effort(-1).minimaxM3().addText('...').message();
 
 - **Gemini:** Gemini 3+ uses bands 0–24 / 25–49 / 50–74 / 75–100. Gemini 3.7 Flash clamps these bands to `low` / `low` / `medium` / `high`; `-1` leaves its native `medium` default unchanged. Gemini 2.5 maps 0–100 to `thinkingBudget`.
 - **GPT-5.6:** `100` maps to `max`; 80–99 remains `xhigh`.
-- **Qwen 3.8 27B:** 0–39 / 40–79 / 80–100 map to `low` / `medium` / `xhigh`; `-1` leaves the native `xhigh` default unchanged.
-- **GLM 5.3:** reasoning is mandatory; 0–39 / 40–79 / 80–100 map to `low` / `high` / `max`; `-1` leaves the native `max` default unchanged.
+- **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.
 - **DeepSeek:** `↑` means thinking is enabled; `off` means it is disabled.
 - **MiniMax:** `off` maps to `thinking.disabled`; `adaptive` maps to `thinking.type=adaptive`.
 - **Anthropic:** Claude 5, Fable, Opus 4.6+, and Sonnet 4.6+ use adaptive thinking with `output_config.effort`. Sonnet 4.5 and Haiku 4.5 use `thinking.type=enabled` with `budget_tokens`.

package/demo/short.js

@@ -16,10 +16,12 @@ const mmix = await ModelMix.new(setup)
     .grok46() // (fallback 4) Grok grok-4.6
     .qwen35397b() // (fallback 5) OpenRouter qwen/qwen3.5-397b-a17b
     .qwen3827b() // (fallback 6) OpenRouter qwen/qwen3.8-27b
-    .GLM53() // (fallback 7) OpenRouter z-ai/glm-5.3
-    .museGlimmer30b({ mix: { fireworks: false, openrouter: true } }) // (fallback 8) OpenRouter meta/muse-glimmer-30b
-    .hermes470b() // (fallback 9) OpenRouter nousresearch/hermes-4-70b
-    .hermes4405b() // (fallback 10) OpenRouter nousresearch/hermes-4-405b
+    .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
     .addText("What's your name?");
 
 console.log(await mmix.message());

package/effort.js

@@ -50,7 +50,9 @@ const OPENAI_MODEL_LEVELS = {
     'meta-models/Muse-Glimmer-30B': ['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'],
     'z-ai/glm-5.3': ['low', 'high', 'max'],
+    'z-ai/glm-5.3-flash': ['low', 'high', 'max'],
     'grok-4.6': ['low', 'medium', 'high', 'xhigh'],
     'gpt-5': ['minimal', 'low', 'medium', 'high'],
     'gpt-5-mini': ['minimal', 'low', 'medium', 'high'],
@@ -396,7 +398,7 @@ function mapEffort(providerFamily, effort, modelKey) {
             return null;
         }
         const supported = supportedOpenAILevels(modelKey);
-        const desired = modelKey === 'z-ai/glm-5.3'
+        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'

package/index.d.ts

@@ -502,6 +502,7 @@ export declare class ModelMix {
   qwen37plus(args?: ModelAttachArgs): this;
   qwen38max(args?: ModelAttachArgs): this;
   qwen3827b(args?: ModelAttachArgs): this;
+  qwen38flash(args?: ModelAttachArgs): this;
   hermes470b(args?: ModelAttachArgs): this;
   hermes4405b(args?: ModelAttachArgs): this;
   hermes3(args?: ModelAttachArgs): this;
@@ -518,6 +519,7 @@ export declare class ModelMix {
   deepseekV4Flash(args?: ModelAttachArgs): this;
   GLM52(args?: ModelAttachArgs): this;
   GLM53(args?: ModelAttachArgs): this;
+  GLM53Flash(args?: ModelAttachArgs): this;
 
   addText(text: string, options?: RoleOptions): this;
   addTextFromFile(filePath: string, options?: RoleOptions): this;

package/index.js

@@ -616,6 +616,10 @@ class ModelMix {
         return this.attach('qwen/qwen3.8-27b', new MixOpenRouter({ options, config }));
     }
 
+    qwen38flash({ options = {}, config = {} } = {}) {
+        return this.attach('qwen/qwen3.8-flash', new MixOpenRouter({ options, config }));
+    }
+
     hermes470b({ options = {}, config = {} } = {}) {
         return this.attach('nousresearch/hermes-4-70b', new MixOpenRouter({ options, config }));
     }
@@ -732,6 +736,10 @@ class ModelMix {
         return this.attach('z-ai/glm-5.3', new MixOpenRouter({ options, config }));
     }
 
+    GLM53Flash({ options = {}, config = {} } = {}) {
+        return this.attach('z-ai/glm-5.3-flash', new MixOpenRouter({ options, config }));
+    }
+
     addText(text, { role = "user", cache } = {}) {
         return this._addText(text, {
             role,

package/lib/model-chain.js

@@ -11,11 +11,11 @@ const CHAIN_MODEL_SHORTCUTS = new Set([
     'gemini35flashLite', 'gemini31flashLite', 'sonarPro', 'sonar',
     'grok46', 'grok45', 'grok43', 'grok420multiAgent', 'grok420',
     'museGlimmer30b',
-    'qwen35397b', 'qwen36plus', 'qwen37plus', 'qwen38max', 'qwen3827b',
+    'qwen35397b', 'qwen36plus', 'qwen37plus', 'qwen38max', 'qwen3827b', 'qwen38flash',
     'hermes470b', 'hermes4405b', 'hermes3',
     'kimiK26', 'kimiK27Code', 'kimiK3', 'kimiK25',
     'minimaxM27', 'minimaxM3', 'mimo25', 'mimo25pro',
-    'deepseekV4Pro', 'deepseekV4Flash', 'GLM52', 'GLM53'
+    'deepseekV4Pro', 'deepseekV4Flash', 'GLM52', 'GLM53', 'GLM53Flash'
 ]);
 
 function parseChainModels(modelSpecs) {

package/lib/token-usage.js

@@ -91,6 +91,7 @@ const MODEL_PRICING = {
     'accounts/fireworks/models/glm-5p2': { input: 1.40, cachedInput: 0.14, output: 4.40 },
     'z-ai/glm-5.2': { input: 0.966, cachedInput: 0.1932, output: 3.036 },
     'z-ai/glm-5.3': { input: 1.40, cachedInput: 0.26, output: 4.40 },
+    'z-ai/glm-5.3-flash': { input: 0.075, cachedInput: 0.015, output: 0.25 },
     'accounts/fireworks/models/kimi-k2p5': { input: 0.50, output: 2.80 },
     'qwen/qwen3.5-397b-a17b': { input: 0.385, output: 2.45 },
     'accounts/fireworks/models/qwen3p6-plus': { input: 0.50, output: 3.00 },
@@ -113,6 +114,7 @@ const MODEL_PRICING = {
     'accounts/fireworks/models/qwen3p8-2p4t-a95b': { input: 2.00, cachedInput: 0.25, output: 6.00 },
     'qwen/qwen3.8-max': { input: 2.00, output: 6.00 },
     'qwen/qwen3.8-27b': { input: 0.45, cachedInput: 0.05, output: 3.20 },
+    'qwen/qwen3.8-flash': { input: 0.16, cachedInput: 0.016, cacheWrite: 0.20, output: 0.47 },
     // MiniMax
     'MiniMax-M2.5': { input: 0.30, output: 1.20 },
     'MiniMax-M2.7': { input: 0.30, output: 1.20 },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "modelmix",
-  "version": "5.1.8",
+  "version": "5.1.9",
   "description": "🧬 Reliable interface with automatic fallback for AI LLMs.",
   "main": "index.js",
   "types": "index.d.ts",

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 maps 0–39 / 40–79 / 80–100 to `low` / `medium` / `xhigh`. GLM 5.3 requires reasoning and maps those bands to `low` / `high` / `max`. Gemini bands: 0–24 / 25–49 / 50–74 / 75–100. Gemini 3.7 Flash supports only `low` / `medium` / `high`, so the first two bands clamp to `low`; `-1` keeps its 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-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 3.7 Flash supports only `low` / `medium` / `high`, so the first two bands clamp to `low`; `-1` keeps its 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
 
@@ -186,7 +186,7 @@ Use `.effort(n)` (or `config.effort`) to enable Anthropic thinking — e.g. `.ef
 `GLM46()`
 
 ### OpenRouter
-`museGlimmer30b()` `gptOss()` `qwen35397b()` `qwen36plus()` `qwen37plus()` `qwen3827b()` `hermes470b()` `hermes4405b()` `qwen38max()` `kimiK27Code()` `kimiK3()` `minimaxM27()` `minimaxM3()` `GLM45()` `GLM52()` `GLM53()`
+`museGlimmer30b()` `gptOss()` `qwen35397b()` `qwen36plus()` `qwen37plus()` `qwen3827b()` `qwen38flash()` `hermes470b()` `hermes4405b()` `qwen38max()` `kimiK27Code()` `kimiK3()` `minimaxM27()` `minimaxM3()` `GLM45()` `GLM52()` `GLM53()` `GLM53Flash()`
 
 ### Multi-provider (auto-fallback across free/paid tiers)
 `hermes3()` `kimiK25()`

package/test/effort.test.js

@@ -140,6 +140,14 @@ describe('Unified effort scale', () => {
             expect(mapEffort('openai', -1, key)).to.equal(null);
         });
 
+        it('maps Qwen 3.8 Flash to its supported reasoning levels', () => {
+            const key = 'qwen/qwen3.8-flash';
+            expect(mapEffort('openai', 0, key)).to.deep.equal({ reasoning_effort: 'low' });
+            expect(mapEffort('openai', 50, key)).to.deep.equal({ reasoning_effort: 'medium' });
+            expect(mapEffort('openai', 100, key)).to.deep.equal({ reasoning_effort: 'xhigh' });
+            expect(mapEffort('openai', -1, key)).to.equal(null);
+        });
+
         it('maps GLM 5.3 to mandatory low, high, and max reasoning', () => {
             const key = 'z-ai/glm-5.3';
             expect(mapEffort('openai', 39, key)).to.deep.equal({ reasoning_effort: 'low' });
@@ -149,6 +157,15 @@ describe('Unified effort scale', () => {
             expect(mapEffort('openai', -1, key)).to.equal(null);
         });
 
+        it('maps GLM 5.3 Flash to mandatory low, high, and max reasoning', () => {
+            const key = 'z-ai/glm-5.3-flash';
+            expect(mapEffort('openai', 39, key)).to.deep.equal({ reasoning_effort: 'low' });
+            expect(mapEffort('openai', 40, key)).to.deep.equal({ reasoning_effort: 'high' });
+            expect(mapEffort('openai', 79, key)).to.deep.equal({ reasoning_effort: 'high' });
+            expect(mapEffort('openai', 80, key)).to.deep.equal({ reasoning_effort: 'max' });
+            expect(mapEffort('openai', -1, key)).to.equal(null);
+        });
+
         it('maps Anthropic adaptive models to thinking + output_config.effort', () => {
             expect(mapEffort('anthropic', 10, 'claude-opus-5')).to.deep.equal({
                 thinking: { type: 'adaptive', display: 'summarized' },

package/test/glm.test.js

@@ -53,4 +53,24 @@ describe('GLM Model Registration Tests', () => {
         expect(model.models).to.have.length(1);
         expect(model.models[0].key).to.equal('z-ai/glm-5.3');
     });
+
+    it('should register GLM 5.3 Flash through OpenRouter', () => {
+        const model = ModelMix.new().GLM53Flash();
+
+        expect(model.models).to.have.length(1);
+        expect(model.models[0].key).to.equal('z-ai/glm-5.3-flash');
+        expect(model.models[0].provider).to.be.instanceOf(MixOpenRouter);
+        expect(ModelMix.calculateCost('z-ai/glm-5.3-flash', {
+            input: 1_000_000,
+            cached: 500_000,
+            output: 1_000_000
+        })).to.equal(0.295);
+    });
+
+    it('should support GLM 5.3 Flash in chain()', () => {
+        const model = ModelMix.new().chain('GLM53Flash');
+
+        expect(model.models).to.have.length(1);
+        expect(model.models[0].key).to.equal('z-ai/glm-5.3-flash');
+    });
 });

package/test/qwen.test.js

@@ -110,6 +110,27 @@ describe('Qwen Model Registration Tests', () => {
         expect(model.models[0].key).to.equal('qwen/qwen3.8-27b');
     });
 
+    it('should register Qwen 3.8 Flash through OpenRouter', () => {
+        const model = ModelMix.new().qwen38flash();
+
+        expect(model.models).to.have.length(1);
+        expect(model.models[0].key).to.equal('qwen/qwen3.8-flash');
+        expect(model.models[0].provider).to.be.instanceOf(MixOpenRouter);
+        expect(ModelMix.calculateCost('qwen/qwen3.8-flash', {
+            input: 1_000_000,
+            cached: 250_000,
+            cacheWrite: 100_000,
+            output: 1_000_000
+        })).to.equal(0.598);
+    });
+
+    it('should support Qwen 3.8 Flash in chain()', () => {
+        const model = ModelMix.new().chain('qwen38flash');
+
+        expect(model.models).to.have.length(1);
+        expect(model.models[0].key).to.equal('qwen/qwen3.8-flash');
+    });
+
     it('should register Qwen 3.5 397B A17B through OpenRouter', () => {
         const model = ModelMix.new().qwen35397b();