From 1e695ba95f1a12c2cb364d06e60050e19aaef8d3 Mon Sep 17 00:00:00 2001 From: Dylan Huang Date: Tue, 27 Jan 2026 10:57:53 -0800 Subject: [PATCH] feat(fireworks): sync model catalog with latest API response - Add Kimi K2.5 (vision + reasoning, 262k context) - Add Qwen3 VL 30B A3B Instruct - Update MiniMax models to not be reasoning (agentic/coding instead) - Update display names to match API (Deepseek vs DeepSeek) - Update model count in docs from 24 to 26 --- docs/providers/fireworks.md | 16 +++++++++------- src/agents/fireworks-models.ts | 28 ++++++++++++++++++++++------ 2 files changed, 31 insertions(+), 13 deletions(-) diff --git a/docs/providers/fireworks.md b/docs/providers/fireworks.md index 6719ec84a..a4dc4766f 100644 --- a/docs/providers/fireworks.md +++ b/docs/providers/fireworks.md @@ -22,7 +22,7 @@ Fireworks AI provides fast, cost-effective serverless inference for popular open - **Streaming**: Supported on all models - **Function calling**: Supported on most models - **Vision**: Supported on vision-capable models (Qwen VL series) -- **Reasoning models**: DeepSeek R1, Qwen3 Thinking, Kimi K2 Thinking +- **Reasoning models**: DeepSeek R1, Qwen3 Thinking, Kimi K2 Thinking, Kimi K2.5 ## Setup @@ -100,26 +100,27 @@ clawdbot models list | grep fireworks | **Fast + cheap** | `qwen3-8b` | Lightweight, low latency | | **Long context** | `kimi-k2-instruct-0905` | 262k context window | -## Available Models (24 Total) +## Available Models (26 Total) ### Text Models | Model ID | Name | Context | Features | |----------|------|---------|----------| -| `deepseek-r1-0528` | DeepSeek R1 05/28 | 163k | Reasoning | -| `deepseek-v3-0324` | DeepSeek V3 03-24 | 163k | General | +| `deepseek-r1-0528` | Deepseek R1 05/28 | 163k | Reasoning | +| `deepseek-v3-0324` | Deepseek V3 03-24 | 163k | General | | `deepseek-v3p1` | DeepSeek V3.1 | 163k | General | | `deepseek-v3p1-terminus` | DeepSeek V3.1 Terminus | 163k | General | -| `deepseek-v3p2` | DeepSeek V3.2 | 163k | General | +| `deepseek-v3p2` | Deepseek v3.2 | 163k | General | | `glm-4p6` | GLM-4.6 | 202k | Reasoning | | `glm-4p7` | GLM-4.7 | 202k | Reasoning | | `gpt-oss-120b` | OpenAI gpt-oss-120b | 131k | Reasoning | | `gpt-oss-20b` | OpenAI gpt-oss-20b | 131k | Reasoning | | `kimi-k2-instruct-0905` | Kimi K2 Instruct 0905 | 262k | Long context | +| `kimi-k2p5` | Kimi K2.5 | 262k | Vision, reasoning | | `kimi-k2-thinking` | Kimi K2 Thinking | 256k | Reasoning | | `llama-v3p3-70b-instruct` | Llama 3.3 70B Instruct | 131k | General | -| `minimax-m2` | MiniMax-M2 | 196k | Reasoning | -| `minimax-m2p1` | MiniMax-M2.1 | 204k | Reasoning | +| `minimax-m2` | MiniMax-M2 | 196k | Agentic, coding | +| `minimax-m2p1` | MiniMax-M2.1 | 204k | Agentic, coding | | `qwen3-235b-a22b` | Qwen3 235B A22B | 131k | General | | `qwen3-235b-a22b-instruct-2507` | Qwen3 235B A22B Instruct 2507 | 262k | General | | `qwen3-235b-a22b-thinking-2507` | Qwen3 235B A22B Thinking 2507 | 262k | Reasoning | @@ -133,6 +134,7 @@ clawdbot models list | grep fireworks | `qwen2p5-vl-32b-instruct` | Qwen2.5-VL 32B Instruct | 128k | Vision | | `qwen3-vl-235b-a22b-instruct` | Qwen3 VL 235B A22B Instruct | 262k | Vision | | `qwen3-vl-235b-a22b-thinking` | Qwen3 VL 235B A22B Thinking | 262k | Vision, reasoning | +| `qwen3-vl-30b-a3b-instruct` | Qwen3 VL 30B A3B Instruct | 262k | Vision | | `qwen3-vl-30b-a3b-thinking` | Qwen3 VL 30B A3B Thinking | 262k | Vision, reasoning | ## Model Catalog diff --git a/src/agents/fireworks-models.ts b/src/agents/fireworks-models.ts index 556a5d70a..4ca2e2132 100644 --- a/src/agents/fireworks-models.ts +++ b/src/agents/fireworks-models.ts @@ -25,7 +25,7 @@ export const FIREWORKS_MODEL_CATALOG = [ // DeepSeek models { id: "accounts/fireworks/models/deepseek-r1-0528", - name: "DeepSeek R1 05/28", + name: "Deepseek R1 05/28", reasoning: true, input: ["text"] as const, contextWindow: 163840, @@ -33,7 +33,7 @@ export const FIREWORKS_MODEL_CATALOG = [ }, { id: "accounts/fireworks/models/deepseek-v3-0324", - name: "DeepSeek V3 03-24", + name: "Deepseek V3 03-24", reasoning: false, input: ["text"] as const, contextWindow: 163840, @@ -57,7 +57,7 @@ export const FIREWORKS_MODEL_CATALOG = [ }, { id: "accounts/fireworks/models/deepseek-v3p2", - name: "DeepSeek V3.2", + name: "Deepseek v3.2", reasoning: false, input: ["text"] as const, contextWindow: 163840, @@ -109,6 +109,14 @@ export const FIREWORKS_MODEL_CATALOG = [ contextWindow: 262144, maxTokens: 8192, }, + { + id: "accounts/fireworks/models/kimi-k2p5", + name: "Kimi K2.5", + reasoning: true, + input: ["text", "image"] as const, + contextWindow: 262144, + maxTokens: 8192, + }, { id: "accounts/fireworks/models/kimi-k2-thinking", name: "Kimi K2 Thinking", @@ -129,11 +137,11 @@ export const FIREWORKS_MODEL_CATALOG = [ maxTokens: 8192, }, - // MiniMax models - "long-thinking" and "complex reasoning tasks" + // MiniMax models - compact MoE for coding and agentic tasks { id: "accounts/fireworks/models/minimax-m2", name: "MiniMax-M2", - reasoning: true, + reasoning: false, input: ["text"] as const, contextWindow: 196608, maxTokens: 8192, @@ -141,7 +149,7 @@ export const FIREWORKS_MODEL_CATALOG = [ { id: "accounts/fireworks/models/minimax-m2p1", name: "MiniMax-M2.1", - reasoning: true, + reasoning: false, input: ["text"] as const, contextWindow: 204800, maxTokens: 8192, @@ -214,6 +222,14 @@ export const FIREWORKS_MODEL_CATALOG = [ contextWindow: 262144, maxTokens: 8192, }, + { + id: "accounts/fireworks/models/qwen3-vl-30b-a3b-instruct", + name: "Qwen3 VL 30B A3B Instruct", + reasoning: false, + input: ["text", "image"] as const, + contextWindow: 262144, + maxTokens: 8192, + }, { id: "accounts/fireworks/models/qwen3-vl-30b-a3b-thinking", name: "Qwen3 VL 30B A3B Thinking",