modelmix 5.1.15 → 5.1.16

package/README.md

@@ -88,7 +88,7 @@ const setup = {
 const model = await ModelMix.new(setup)
     .sonnet5() // (main model) Anthropic claude-sonnet-5
     .gpt56luna() // (fallback 2) OpenAI gpt-5.6-luna
-    .gemini37flash() // (fallback 3) Google gemini-3.7-flash
+    .gemini38flash() // (fallback 3) Google gemini-3.8-flash
     .grok46() // (fallback 4) Grok grok-4.6
     .addText("What's your name?");
 
@@ -122,7 +122,7 @@ has no configured effort:
 
 ```javascript
 const model = ModelMix.new(setup)
-    .chain('sonnet5', 'gpt56luna@20', 'gemini37flash@-1')
+    .chain('sonnet5', 'gpt56luna@20', 'gemini38flash@-1')
     .addText("What's your name?");
 
 console.log(await model.message());
@@ -185,6 +185,7 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `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] |
 | `gemini35flash()` | Google | gemini-3.5-flash | [\$0.75][3] | [\$4.50][3] |
@@ -220,7 +221,7 @@ ModelMix provides convenient shorthand methods for quickly accessing different A
 | `kimiK25()` | Together | Kimi-K2.5 | [\$0.50][7] | [\$2.80][7] |
 | `kimiK26()` | Fireworks | models/kimi-k2p6 | [\$0.95][10] | [\$4.00][10] |
 
-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.
+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.
 
 `museGlimmer30b()` uses Fireworks by default. OpenRouter, NVIDIA NIM, and Together are available through `mix.openrouter`, `mix.nvidia`, and `mix.together`.
 
@@ -291,7 +292,7 @@ ModelMix.new().effort(-1).minimaxM3().addText('...').message();
 
 ### Provider-specific behavior
 
-- **Gemini:** Gemini 3+ uses bands 0–24 / 25–49 / 50–74 / 75–100. Gemini 3.7 Flash clamps these bands to `low` / `low` / `medium` / `high`; `-1` leaves its native `medium` default unchanged. Gemini 2.5 maps 0–100 to `thinkingBudget`.
+- **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-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/gemini.js

@@ -12,9 +12,9 @@ const mmix = new ModelMix({
     }
 });
 
-// Using Gemini 3.7 Flash with the built-in method
-console.log("\n" + '--------| gemini37flash() |--------');
-const flash = await mmix.gemini37flash()
+// Using Gemini 3.8 Flash with the built-in method
+console.log("\n" + '--------| gemini38flash() |--------');
+const flash = await mmix.gemini38flash()
     .addText('Hi there! Do you like cats?')
     .message();
 

package/demo/short.js

@@ -12,7 +12,7 @@ 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
     .gpt56luna({ mix: { openrouter: true } }) // (fallback 3 + provider fallback) OpenAI/OpenRouter gpt-5.6-luna
-    .gemini37flash() // (fallback 4) Google gemini-3.7-flash
+    .gemini38flash() // (fallback 4) Google gemini-3.8-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

package/effort.js

@@ -96,6 +96,7 @@ const MINIMAX_BANDS = [
 
 /** Exact model → supported Gemini thinkingLevel values */
 const GEMINI_MODEL_LEVELS = {
+    'gemini-3.8-flash': ['low', 'medium', 'high'],
     'gemini-3.7-flash': ['low', 'medium', 'high'],
     'gemini-3-pro-preview': ['low', 'high'],
     'gemini-3.1-pro-preview': ['low', 'medium', 'high'],
@@ -320,7 +321,7 @@ function mapAdaptiveEffort(providerFamily, modelKey) {
         return { thinking: { type: 'adaptive' } };
     }
     if (providerFamily === 'google') {
-        if (modelKey === 'gemini-3.7-flash') return null;
+        if (modelKey === 'gemini-3.8-flash' || modelKey === 'gemini-3.7-flash') return null;
         // Gemini dynamic thinking: thinkingBudget -1 (2.5 official; accepted on 3.x as dynamic)
         return { thinkingConfig: { thinkingBudget: -1 } };
     }

package/index.d.ts

@@ -482,6 +482,7 @@ export declare class ModelMix {
 
   // Google
   gemini31pro(args?: ModelAttachArgs): this;
+  gemini38flash(args?: ModelAttachArgs): this;
   gemini37flash(args?: ModelAttachArgs): this;
   gemini36flash(args?: ModelAttachArgs): this;
   gemini35flash(args?: ModelAttachArgs): this;

package/index.js

@@ -573,6 +573,9 @@ class ModelMix {
     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 }));
+    }
     gemini37flash({ options = {}, config = {} } = {}) {
         return this.attach('gemini-3.7-flash', new MixGoogle({ options, config }));
     }

package/lib/model-chain.js

@@ -7,7 +7,7 @@ const CHAIN_MODEL_SHORTCUTS = new Set([
     'gptRealtime', 'gptRealtimeMini', 'gpt53codex', 'gpt53chat', 'gptOss',
     'fable51', 'fable50', 'fable5', 'opus50', 'opus5', 'opus48', 'opus47', 'opus46',
     'sonnet50', 'sonnet5', 'sonnet46', 'sonnet45', 'haiku45',
-    'gemini31pro', 'gemini37flash', 'gemini36flash', 'gemini35flash',
+    'gemini31pro', 'gemini38flash', 'gemini37flash', 'gemini36flash', 'gemini35flash',
     'gemini35flashLite', 'gemini31flashLite', 'sonarPro', 'sonar',
     'grok46', 'grok45', 'grok43', 'grok420multiAgent', 'grok420',
     'museGlimmer30b', 'museSpark12Contributor',

package/lib/token-usage.js

@@ -64,6 +64,7 @@ const MODEL_PRICING = {
     'gemini-3.1-pro-preview': { input: 2.00, output: 12.00 },
     'gemini-3-pro-preview': { input: 2.00, output: 12.00 },
     'gemini-3-flash-preview': { input: 0.50, output: 3.00 },
+    'gemini-3.8-flash': { input: 0.75, cachedInput: 0.075, output: 3.75 },
     'gemini-3.7-flash': { input: 0.75, cachedInput: 0.075, output: 3.75 },
     'gemini-3.6-flash': { input: 0.75, cachedInput: 0.075, output: 3.75 },
     'gemini-3.5-flash': { input: 0.75, output: 4.50 },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "modelmix",
-  "version": "5.1.15",
+  "version": "5.1.16",
   "description": "🧬 Reliable interface with automatic fallback for AI LLMs.",
   "main": "index.js",
   "types": "index.d.ts",

package/skills/modelmix/SKILL.md

@@ -86,11 +86,11 @@ Chain shorthand methods to attach providers. First model is primary; others are
 const model = ModelMix.new()
     .sonnet46()        // primary
     .gpt52()           // fallback 1
-    .gemini37flash()   // fallback 2
+    .gemini38flash()   // fallback 2
     .addText("Hello!")
 ```
 
-If `sonnet46` fails, it automatically tries `gpt52`, then `gemini37flash`.
+If `sonnet46` fails, it automatically tries `gpt52`, then `gemini38flash`.
 
 The equivalent `chain()` form accepts public shortcut names directly in the
 same order. Append `@effort` for a per-model unified effort override (`-1` or
@@ -99,7 +99,7 @@ provider default when no chain effort is configured:
 
 ```javascript
 const model = ModelMix.new()
-    .chain('sonnet46', 'gpt52@20', 'gemini37flash@-1')
+    .chain('sonnet46', 'gpt52@20', 'gemini38flash@-1')
     .addText('Hello!');
 ```
 
@@ -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 3.7 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 3.8 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
 
@@ -163,7 +163,7 @@ Use `.effort(n)` (or `config.effort`) to enable Anthropic thinking — e.g. `.ef
 `fable51()` registers `claude-fable-5-1` through Anthropic by default. Pass `mix: { openrouter: true }` to append `anthropic/claude-fable-5.1` as its fallback.
 
 ### Google
-`gemini31pro()` `gemini37flash()` `gemini36flash()` `gemini35flash()` `gemini35flashLite()` `gemini31flashLite()`
+`gemini31pro()` `gemini38flash()` `gemini37flash()` `gemini36flash()` `gemini35flash()` `gemini35flashLite()` `gemini31flashLite()`
 
 ### Grok
 `grok46()` `grok45()` `grok43()` `grok420multiAgent()` `grok420()`
@@ -485,7 +485,7 @@ Omit all weights for equal probabilities. Otherwise every option needs a positiv
 const pool = ModelMix.new({ config: { roundRobin: true } })
     .gpt5mini()
     .sonnet45()
-    .gemini37flash();
+    .gemini38flash();
 
 const r1 = await pool.new().addText("Request 1").message();
 const r2 = await pool.new().addText("Request 2").message();

package/test/effort.test.js

@@ -242,6 +242,19 @@ describe('Unified effort scale', () => {
             expect(mapEffort('google', -1, 'gemini-3.7-flash')).to.equal(null);
         });
 
+        it('clamps Gemini 3.8 Flash to low, medium, and high', () => {
+            expect(mapEffort('google', 0, 'gemini-3.8-flash')).to.deep.equal({
+                thinkingConfig: { thinkingLevel: 'low' }
+            });
+            expect(mapEffort('google', 50, 'gemini-3.8-flash')).to.deep.equal({
+                thinkingConfig: { thinkingLevel: 'medium' }
+            });
+            expect(mapEffort('google', 100, 'gemini-3.8-flash')).to.deep.equal({
+                thinkingConfig: { thinkingLevel: 'high' }
+            });
+            expect(mapEffort('google', -1, 'gemini-3.8-flash')).to.equal(null);
+        });
+
         it('clamps Gemini levels for models with fewer steps', () => {
             expect(mapEffort('google', 10, 'gemini-3-pro-preview')).to.deep.equal({
                 thinkingConfig: { thinkingLevel: 'low' }

package/test/fallback.test.js

@@ -34,12 +34,12 @@ describe('Provider Fallback Chain Tests', () => {
         });
 
         it('should attach chain shortcuts from arguments with optional per-model effort', () => {
-            model.chain('sonnet5', 'gpt56luna@20', 'gemini37flash@-1');
+            model.chain('sonnet5', 'gpt56luna@20', 'gemini38flash@-1');
 
             expect(model.models.map(({ key }) => key)).to.deep.equal([
                 'claude-sonnet-5',
                 'gpt-5.6-luna',
-                'gemini-3.7-flash'
+                'gemini-3.8-flash'
             ]);
             expect(model.models[0].provider.config).to.not.have.property('effort');
             expect(model.models[1].provider.config.effort).to.equal(20);

package/test/live.mcp.js

@@ -181,8 +181,8 @@ describe('Live MCP Integration Tests', function () {
             }
         });
 
-        it('should use custom MCP tools with Gemini 3.7 Flash', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+        it('should use custom MCP tools with Gemini 3.8 Flash', async function () {
+            const model = ModelMix.new(setup).gemini38flash();
 
             // Add password generator tool
             model.addTool({
@@ -220,7 +220,7 @@ describe('Live MCP Integration Tests', function () {
             model.addText('Generate a secure password of 16 characters with symbols.');
 
             const response = await model.message();
-            console.log(`Gemini 3.7 Flash with MCP tools: ${response}`);
+            console.log(`Gemini 3.8 Flash with MCP tools: ${response}`);
 
             expect(response).to.be.a('string');
             // Check password is mentioned and a generated password string is present
@@ -407,8 +407,8 @@ describe('Live MCP Integration Tests', function () {
             expect(result.factorial_result).to.equal(120);
         });
 
-        it('should use MCP tools with JSON output using Gemini 3.7 Flash', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+        it('should use MCP tools with JSON output using Gemini 3.8 Flash', async function () {
+            const model = ModelMix.new(setup).gemini38flash();
 
             // Add system info tool
             model.addTool({
@@ -447,7 +447,7 @@ describe('Live MCP Integration Tests', function () {
                 generated_at: ""
             });
 
-            console.log(`Gemini 3.7 Flash with MCP tools JSON result:`, result);
+            console.log(`Gemini 3.8 Flash with MCP tools JSON result:`, result);
 
             expect(result).to.be.an('object');
             expect(result.timestamp).to.be.a('number');

package/test/live.test.js

@@ -59,7 +59,7 @@ describe('Live Integration Tests', function () {
         });
 
         it('should process images with Google Gemini', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+            const model = ModelMix.new(setup).gemini38flash();
 
             model.addImageFromUrl(blueSquareBase64)
                 .addText('What color is this image? Answer in one word only.');
@@ -120,7 +120,7 @@ describe('Live Integration Tests', function () {
         });
 
         it('should return structured JSON with Google Gemini', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+            const model = ModelMix.new(setup).gemini38flash();
 
             model.addText('Generate information about a fictional city.');
 

package/test/tokens.test.js

@@ -517,26 +517,29 @@ describe('Token Usage Tracking', () => {
 
     it('should register Gemini Flash shortcuts with Google provider', function () {
         const model = ModelMix.new()
+            .gemini38flash()
             .gemini37flash()
             .gemini36flash()
             .gemini35flash()
             .gemini35flashLite();
 
         expect(model.models.map(({ key }) => key)).to.deep.equal([
+            'gemini-3.8-flash',
             'gemini-3.7-flash',
             'gemini-3.6-flash',
             'gemini-3.5-flash',
             'gemini-3.5-flash-lite'
         ]);
         expect(model.models.every(({ provider }) => provider instanceof MixGoogle)).to.equal(true);
+        expect(ModelMix.calculateCost('gemini-3.8-flash', { input: 1_000_000, output: 1_000_000 })).to.equal(4.5);
         expect(ModelMix.calculateCost('gemini-3.7-flash', { input: 1_000_000, output: 1_000_000 })).to.equal(4.5);
         expect(ModelMix.calculateCost('gemini-3.6-flash', { input: 1_000_000, output: 1_000_000 })).to.equal(4.5);
         expect(ModelMix.calculateCost('gemini-3.5-flash', { input: 1_000_000, output: 1_000_000 })).to.equal(5.25);
         expect(ModelMix.calculateCost('gemini-3.5-flash-lite', { input: 1_000_000, output: 1_000_000 })).to.equal(2.8);
     });
 
-    it('should calculate Gemini 3.7 Flash cache reads at the introductory rate', function () {
-        expect(ModelMix.calculateCostBreakdown('gemini-3.7-flash', {
+    it('should calculate Gemini 3.8 Flash cache reads and thinking at the introductory rate', function () {
+        expect(ModelMix.calculateCostBreakdown('gemini-3.8-flash', {
             input: 1_000_000,
             output: 1_000_000,
             thinking: 500_000,
@@ -552,10 +555,10 @@ describe('Token Usage Tracking', () => {
         });
     });
 
-    it('should forward options and config through gemini37flash()', function () {
+    it('should forward options and config through gemini38flash()', function () {
         const options = { thinkingLevel: 'high' };
         const config = { max_history: 3 };
-        const model = ModelMix.new().gemini37flash({ options, config });
+        const model = ModelMix.new().gemini38flash({ options, config });
 
         expect(model.models[0].provider).to.be.instanceOf(MixGoogle);
         expect(model.models[0].provider.options).to.deep.equal(options);
@@ -687,7 +690,7 @@ describe('Token Usage Tracking', () => {
         this.timeout(30000);
 
         const model = ModelMix.new()
-            .gemini37flash()
+            .gemini38flash()
             .addText('Say hi');
 
         const result = await model.raw();
@@ -750,7 +753,7 @@ describe('Token Usage Tracking', () => {
         const providers = [
             { name: 'OpenAI', create: (m) => m.gpt56luna() },
             { name: 'Anthropic', create: (m) => m.haiku45() },
-            { name: 'Google', create: (m) => m.gemini37flash() }
+            { name: 'Google', create: (m) => m.gemini38flash() }
         ];
 
         for (const provider of providers) {