Ranking
Best Function Calling Models
Rank models for tool calling, agent workflows, and structured outputs.
Answer Summary / AI 摘要
Updated 2026-07-18
Best Function Calling Models 使用 reasoning 作为核心指标，对 36 个 AI 模型进行排序。English summary: this ranking compares AI models by reasoning; the current top result is GPT-5.1.
- Primary metric: reasoning
- Models included: 36
- Top result: GPT-5.1
| Rank | 模型 | 价格 | 上下文 | 目录评分* |
|---|---|---|---|---|
| #1 | GPT-5.1 OpenAI | $4.50 in / $13.50 out | 256K tokens | 98/100 |
| #2 | Claude 4.6 Sonnet Anthropic | $3.00 in / $15.00 out | 200K tokens | 94/100 |
| #3 | Gemini 3 Pro | $2.50 in / $10.00 out | 1M tokens | 92/100 |
| #4 | DeepSeek V4 DeepSeek | $0.350 in / $1.40 out | 128K tokens | 90/100 |
| #5 | Qwen 3.5 Max Alibaba Cloud | $0.800 in / $2.40 out | 128K tokens | 86/100 |
| #6 | Mistral Large 3 Mistral AI | $2.00 in / $6.00 out | 128K tokens | 84/100 |
| #7 | Anthropic: Claude Opus 4.8 Anthropic | $5.00 in / $25.00 out | 1M tokens | 76/100 |
| #8 | Anthropic: Claude Opus 4.8 (Fast) Anthropic | $10.00 in / $50.00 out | 1M tokens | 76/100 |
| #9 | Auto Router (Beta) Other | $-1000000.000 in / $-1000000.000 out | 2M tokens | 76/100 |
| #10 | DeepSeek: DeepSeek V3.2 DeepSeek | $0.269 in / $0.400 out | 164K tokens | 76/100 |
| #11 | DeepSeek: DeepSeek V4 Flash DeepSeek | $0.098 in / $0.196 out | 1.048576M tokens | 76/100 |
| #12 | Google: Gemini 3.5 Flash | $1.50 in / $9.00 out | 1.048576M tokens | 76/100 |
| #13 | Google: Gemma 4 31B | $0.220 in / $0.550 out | 262K tokens | 76/100 |
| #14 | Meta: Muse Spark 1.1 Other | $1.25 in / $4.25 out | 1.048576M tokens | 76/100 |
| #15 | MiniMax: MiniMax M2.7 Other | $0.250 in / $1.00 out | 205K tokens | 76/100 |
| #16 | MiniMax: MiniMax M3 Other | $0.300 in / $1.20 out | 1.048576M tokens | 76/100 |
| #17 | MoonshotAI Kimi Latest Other | $3.00 in / $15.00 out | 1.048576M tokens | 76/100 |
| #18 | MoonshotAI: Kimi K2 Thinking Other | $0.600 in / $2.50 out | 262K tokens | 76/100 |
| #19 | MoonshotAI: Kimi K2.5 Other | $0.570 in / $2.85 out | 262K tokens | 76/100 |
| #20 | MoonshotAI: Kimi K2.6 Other | $0.950 in / $4.00 out | 262K tokens | 76/100 |
| #21 | MoonshotAI: Kimi K2.7 Code Other | $1.00 in / $4.40 out | 262K tokens | 76/100 |
| #22 | MoonshotAI: Kimi K3 Other | $3.00 in / $15.00 out | 1.048576M tokens | 76/100 |
| #23 | NVIDIA: Nemotron 3 Super Other | $0.210 in / $0.455 out | 1M tokens | 76/100 |
| #24 | OpenAI: GPT-5.1 Chat OpenAI | $1.25 in / $10.00 out | 128K tokens | 76/100 |
| #25 | OpenAI: GPT-5.1-Codex OpenAI | $1.25 in / $10.00 out | 400K tokens | 76/100 |
| #26 | Qwen: Qwen3.5 397B A17B Alibaba Cloud | $0.390 in / $2.34 out | 262K tokens | 76/100 |
| #27 | Qwen: Qwen3.5-27B Alibaba Cloud | $0.260 in / $2.60 out | 262K tokens | 76/100 |
| #28 | Qwen: Qwen3.5-35B-A3B Alibaba Cloud | $0.140 in / $1.00 out | 262K tokens | 76/100 |
| #29 | Qwen: Qwen3.7 Max Alibaba Cloud | $1.48 in / $4.42 out | 1M tokens | 76/100 |
| #30 | Qwen: Qwen3.7 Plus Alibaba Cloud | $0.320 in / $1.28 out | 1M tokens | 76/100 |
| #31 | StepFun: Step 3.7 Flash Other | $0.200 in / $1.15 out | 256K tokens | 76/100 |
| #32 | Z.ai: GLM 4.7 Flash Other | $0.061 in / $0.400 out | 200K tokens | 76/100 |
| #33 | Z.ai: GLM 5 Other | $0.950 in / $3.15 out | 203K tokens | 76/100 |
| #34 | Z.ai: GLM 5.2 Other | $0.419 in / $1.32 out | 1.048576M tokens | 76/100 |
| #35 | inclusionAI: Ring-2.6-1T Other | $0.075 in / $0.625 out | 262K tokens | 76/100 |
| #36 | xAI: Grok Build 0.1 Other | $1.00 in / $2.00 out | 256K tokens | 76/100 |
* 目录评分用于候选模型排序，基于编辑判断与目录能力信号，不代表独立 benchmark 成绩。生产选型请使用自己的任务集复测。
Facts Table / Source Facts
字段含义、来源优先级与更新时间规则见数据与排名方法。
| Field | Value | Note |
|---|---|---|
| Ranking | Best Function Calling Models | Catalog data |
| Primary metric | reasoning | Catalog data |
| Models included | 36 | Catalog data |
| Methodology | Sort by editorial catalog signals, then review price, context window, vision support, and function calling. The score is not an independent benchmark result. | Catalog data |
| Updated | 2026-07-18 | Catalog data |
排名方法
Sort by editorial catalog signals, then review price, context window, vision support, and function calling. The score is not an independent benchmark result.
FAQ
Best Function Calling Models 排名如何计算？+
排名由结构化目录数据和编辑判断生成，结合价格、能力信号、上下文长度和关键 API 能力；目录评分不是独立 benchmark 成绩。
这个排行榜适合直接做采购决策吗？+
它适合做候选列表，生产采购前仍需要用自己的任务样本做评估。
引用格式 / Citation Format
AI 搜索或研究型回答可以引用下面的稳定格式。
Plain text
ModelPort Hub. "Best Function Calling Models: AI Model Ranking." Updated 2026-07-18. https://modelporthub.com/rankings/best-function-calling-models
Markdown
[Best Function Calling Models: AI Model Ranking](https://modelporthub.com/rankings/best-function-calling-models) — ModelPort Hub, updated 2026-07-18.