Best AI models for tool use and agents
τ²-Bench measures multi-turn agentic tool use: calling functions, following policies, and completing realistic tasks over many turns. If you are building agents or tool-calling workflows, this predicts real-world reliability better than single-shot benchmarks.
- 🥇
glm-5.2Z.ai·$1.40 / $4.40 per 1M99.1%99.1% - 🥈
claude-fable-5Anthropic·$10.00 / $50.00 per 1M98.5%98.5% - 🥉
stepfun/step-3.7-flashNovita AI·$0.20 / $1.15 per 1M98.5%98.5% - 4GLM-5Z.ai·$1.00 / $3.20 per 1M98.2%98.2%
- 5grok-4.3xAI·$1.25 / $2.50 per 1M97.7%97.7%
- 6glm-5.1Z.ai·$1.40 / $4.40 per 1M97.7%97.7%
- 7
qwen3.6-plusAlibaba Cloud·$0.50 / $3.00 per 1M97.7%97.7% - 8
deepseek-v4-pro-0424DeepInfra·$1.30 / $2.60 per 1M96.2%96.2% - 9
kimi-k2.6Moonshot AI·$0.95 / $4.00 per 1M95.9%95.9% - 10kimi-k2.5Moonshot AI·$0.60 / $3.00 per 1M95.9%95.9%
- 11GLM-4.7Z.ai·$0.60 / $2.20 per 1M95.9%95.9%
- 12
gemini-3.5-flashGoogle Gemini API·$1.50 / $9.00 per 1M95.6%95.6% - 13gemini-3.1-pro-previewGoogle Gemini API·$2.00 / $12.00 per 1M95.6%95.6%
- 14minimax-m2.5MiniMax·$0.30 / $1.20 per 1M95.3%95.3%
- 15qwen3.7-maxAlibaba Cloud·$2.50 / $7.50 per 1M94.7%94.7%
- 16claude-opus-4-8Anthropic·$5.00 / $25.00 per 1M94.4%94.4%
- 17
mistral-medium-3-5Mistral AI·$1.65 / $8.25 per 1M94.2%94.2% - 18XiaomiMiMo/MiMo-V2.5-ProDeepInfra·$1.00 / $3.00 per 1M94.2%94.2%
- 19
gpt-5.5OpenAI·$5.00 / $30.00 per 1M93.9%93.9% - 20Qwen/Qwen3.5-27BDeepInfra·$0.26 / $2.60 per 1M93.9%93.9%
- 21qwen3.7-plus20% offAlibaba Cloud·$0.32 / $1.28 per 1M93.0%93.0%
- 22
kimi-k2Google Vertex AI·$0.60 / $2.50 per 1M93.0%93.0% - 23inclusionai/ring-2.6-1tNovita AI·$0.30 / $2.50 per 1M92.4%92.4%
- 24qwen3.5-4bRunware·$0.05 / $0.07 per 1M92.1%92.1%
- 25claude-opus-4-6Anthropic·$5.00 / $25.00 per 1M92.1%92.1%
- 26
gpt-5.2-codex@eastus2Microsoft Azure AI·$1.75 / $14.00 per 1M92.1%92.1% - 27deepseek-v3.2Google Vertex AI·$0.56 / $1.68 per 1M90.6%90.6%
- 28grok-3-minixAI·$0.30 / $0.50 per 1M90.4%90.4%
- 29kimi-k2.7-codeMoonshot AI·$0.95 / $4.00 per 1M90.1%90.1%
- 30inclusionai/ling-2.6-1tNovita AI·$0.30 / $2.50 per 1M89.8%89.8%
Explore other rankings
How we rank
Scores for τ²-Bench come from Artificial Analysis, an independent AI benchmarking service. When a model is available through multiple providers (e.g. Anthropic direct, AWS Bedrock, Google Vertex), we show one canonical entry per model family so the ranking isn't polluted by duplicates. Benchmarks measure specific skills, so always validate on your own workload before committing.
One API for every model on this list
Requesty is OpenAI-compatible and routes to 600+ models. Switch between any of the models above by changing one parameter in your code.