Best LLMs for Tool Calling — OpenAI
The best OpenAI model for tool calling is GPT-5.5, completing 93.9% of Tau²-Bench's multi-turn tool-use tasks. GPT-5.2-Codex (92.1%) and GPT-5.4 (87.1%) round out the top three.
- 1
gpt-5.5ReasoningToolsJSON+154.8 intel · $5.00/M · 1.1M ctx93.9%τ²-Bench - 2gpt-5.2-codexReasoningToolsJSON+140.1 intel · $1.75/M · 400K ctx92.1%τ²-Bench
- 3gpt-5.4ReasoningToolsJSON+151.4 intel · $2.50/M · 1.1M ctx87.1%τ²-Bench
- 4gpt-5.6-terraReasoningToolsJSON+155.0 intel · $1.00/M · 83 t/s86.3%τ²-Bench
- 5gpt-5.3-codexReasoningToolsJSON+144.3 intel · $1.75/M · 122 t/s86.0%τ²-Bench
- 6gpt-5.6-solReasoningToolsJSON+158.9 intel · $5.00/M · 51 t/s85.1%τ²-Bench
- 7gpt-5.2ReasoningToolsJSON+142.2 intel · $1.75/M · 81 t/s84.8%τ²-Bench
- 8gpt-5ReasoningToolsJSON+134.7 intel · $1.25/M · 65 t/s84.8%τ²-Bench
- 9gpt-5.4-miniReasoningToolsJSON+140.0 intel · $0.750/M · 400K ctx83.3%τ²-Bench
- 10gpt-5.1-codexReasoningToolsJSON+134.7 intel · $1.25/M · 65 t/s83.0%τ²-Bench
- 11gpt-5.1ReasoningToolsJSON+136.9 intel · $1.25/M · 59 t/s81.9%τ²-Bench
- 12o3ReasoningToolsJSON+130.4 intel · $2.00/M · 54 t/s80.7%τ²-Bench
- 13gpt-5.4-nanoReasoningToolsJSON+138.2 intel · $0.200/M · 400K ctx76.0%τ²-Bench
- 14gpt-5-miniReasoningToolsJSON+125.3 intel · $0.250/M · 93 t/s68.4%τ²-Bench
- 15gpt-oss-120bReasoningToolsJSON23.8 intel · $0.037/M · 624 t/s65.8%τ²-Bench
- 16gpt-5.1-codex-miniReasoningToolsJSON+130.6 intel · $0.250/M · 61 t/s62.9%τ²-Bench
- 17o1ReasoningToolsJSON+123.4 intel · $15.00/M · 200K ctx62.6%τ²-Bench
- 18gpt-oss-20bReasoningToolsJSON14.9 intel · $0.030/M · 236 t/s60.2%τ²-Bench
- 19gpt-oss-20b:freeReasoningToolsJSON14.9 intel · Free/M · 17 t/s60.2%τ²-Bench
- 20o4-mini-highReasoningToolsJSON+125.6 intel · $1.10/M · 84 t/s55.6%τ²-Bench
- 21o4-miniReasoningToolsJSON+125.6 intel · $1.10/M · 59 t/s55.6%τ²-Bench
- 22gpt-4.1-miniToolsJSONVision14.8 intel · $0.400/M · 40 t/s52.9%τ²-Bench
- 23gpt-4.1ToolsJSONVision19.4 intel · $2.00/M · 64 t/s47.1%τ²-Bench
- 24gpt-5-nanoReasoningToolsJSON+119.9 intel · $0.050/M · 98 t/s36.5%τ²-Bench
- 25o3-mini-highReasoningToolsJSON15.6 intel · $1.10/M · 200K ctx31.3%τ²-Bench
How this is ranked
AI models ranked by Tau²-Bench — multi-turn conversations where the model has to call the right tools, in the right order, against a real API to complete a customer task. This measures whether function calling actually works under pressure, which is a different question from whether a model supports the parameter at all.
Frequently asked
Which OpenAI model is best at tool calling?
The best OpenAI model for tool calling is GPT-5.5, completing 93.9% of Tau²-Bench's multi-turn tool-use tasks. GPT-5.2-Codex (92.1%) and GPT-5.4 (87.1%) round out the top three.
What's a good alternative to GPT-5.5?
GPT-5.2-Codex (92.1%) is the closest alternative on this metric, followed by GPT-5.4 (87.1%). See the full ranking above for the tradeoffs.
How many OpenAI models are there?
modelgrep tracks 60 OpenAI models with live benchmarks, speed, latency and per-provider pricing, led on intelligence by GPT-5.6 Sol. 25 of them qualify for this ranking.