Benchmark Almanac

Galileo Agent Leaderboard

Claude 3.7 Sonnet 0.95 leads since Feb 2025 Not in model cards

Models pick and call tools on samples drawn from existing tool-use benchmarks. Galileo scores Tool Selection Quality with an LLM judge and averages the datasets equally.

The first release ranked 17 models, 12 closed and 5 open. The Hugging Face space has about 450 likes.

Made by Galileo Released Feb 2025 Function calling All tasks public Last model added Apr 2025 Scores refreshed 2 Nov 2026
Level
Everyday Most adults, without training
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Dataset Leaderboard galileo.ai
33
Entries on its own leaderboard, 31 models we track

Claude 3.7 Sonnet leads with 0.95 among models released by Apr 2025

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

The newest model with a score here came out in Apr 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

33 entries. Scroll sideways for the rest.

  1. 0.95 Claude 3.7 Sonnet
  2. 0.94 Gemini 2.5 Pro
  3. 0.94 Gemini 2.0 Flash
  4. 0.93 Gemini 2.0 Flash-Lite
  5. 0.92 GPT-4.1
  6. 0.91 Mistral Small 3.1
  7. 0.91 GPT-4.1 mini
  8. 0.91 DeepSeek-V3
  9. 0.90 GPT-4o
  10. 0.90 GPT-4.5
  11. 0.90 Gemini 1.5 Flash
  12. 0.89 o4-mini
  13. 0.89 Palmyra X4
  14. 0.89 Gemini 1.5 Pro
  15. 0.88 o1
  16. 0.87 Amazon Nova Lite
  17. 0.87 Amazon Nova Pro
  18. 0.86 o3
  19. 0.85 o3-mini
  20. 0.83 GPT-4o mini
  21. 0.83 Amazon Nova Micro
  22. 0.82 Qwen2.5-72B
  23. 0.81 Mistral Large 2
  24. 0.80 GPT-4.1 nano
  25. 0.80 Claude 3.5 Sonnet (October 2024)
  26. 0.77 Llama-3.3-70B-Instruct-Turbo
  27. 0.77 Claude 3.5 Haiku
  28. 0.75 Mistral Small v24.09
  29. 0.74 Llama 4 Maverick
  30. 0.69 Ministral 8B
  31. 0.68 Meta-Llama-3.1-8B-Instruct-Turbo
  32. 0.66 Mistral NeMo
  33. 0.63 Llama 4 Scout
  1. 1 Claude 3.7 Sonnet 0.95
  2. 2 Gemini 2.5 Pro Mar 2025 0.94
  3. 3 Gemini 2.0 Flash Feb 2025 0.94
  4. 4 Gemini 2.0 Flash-Lite 0.93
  5. 5 GPT-4.1 0.92
  6. 6 Mistral Small 3.1 0.91
  7. 7 GPT-4.1 mini 0.91
  8. 8 DeepSeek-V3 Mar 2025 0.91
  9. 9 GPT-4o Nov 2024 0.90
  10. 10 GPT-4.5 0.90
  11. 11 Gemini 1.5 Flash Sep 2024 0.90
  12. 12 o4-mini 0.89
  13. 13 Palmyra X4 0.89
  14. 14 Gemini 1.5 Pro Sep 2024 0.89
  15. 15 o1 0.88
  16. 16 Amazon Nova Lite 0.87
  17. 17 Amazon Nova Pro 0.87
  18. 18 o3 0.86
  19. 19 o3-mini 0.85
  20. 20 GPT-4o mini 0.83
  21. 21 Amazon Nova Micro 0.83
  22. 22 Qwen2.5-72B 0.82
  23. 23 Mistral Large 2 Nov 2024 0.81
  24. 24 GPT-4.1 nano 0.80
  25. 25 Claude 3.5 Sonnet (October 2024) 0.80
  26. 26 Llama-3.3-70B-Instruct-Turbo 0.77
  27. 27 Claude 3.5 Haiku 0.77
  28. 28 Mistral Small v24.09 0.75
  29. 29 Llama 4 Maverick 0.74
  30. 30 Ministral 8B 0.69
  31. 31 Meta-Llama-3.1-8B-Instruct-Turbo 0.68
  32. 32 Mistral NeMo 0.66
  33. 33 Llama 4 Scout 0.63
#EntryScoreSettingDateSource
1 Claude 3.7 Sonnet 0.95 Feb 2025 Official leaderboard
2 Gemini 2.5 Pro Mar 2025 0.94 Mar 2025 Official leaderboard
3 Gemini 2.0 Flash Feb 2025 0.94 Feb 2025 Official leaderboard
4 Gemini 2.0 Flash-Lite 0.93 Feb 2025 Official leaderboard
5 GPT-4.1 0.92 Apr 2025 Official leaderboard
6 Mistral Small 3.1 0.91 Mar 2025 Official leaderboard
7 GPT-4.1 mini 0.91 Apr 2025 Official leaderboard
8 DeepSeek-V3 Mar 2025 0.91 Mar 2025 Official leaderboard
9 GPT-4o Nov 2024 0.90 Nov 2024 Official leaderboard
10 GPT-4.5 0.90 Feb 2025 Official leaderboard
11 Gemini 1.5 Flash Sep 2024 0.90 Aug 2024 Official leaderboard
12 o4-mini 0.89 Apr 2025 Official leaderboard
13 Palmyra X4 0.89 Oct 2024 Official leaderboard
14 Gemini 1.5 Pro Sep 2024 0.89 Sep 2024 Official leaderboard
15 o1 0.88 Dec 2024 Official leaderboard
16 Amazon Nova Lite 0.87 Dec 2024 Official leaderboard
17 Amazon Nova Pro 0.87 Dec 2024 Official leaderboard
18 o3 0.86 Apr 2025 Official leaderboard
19 o3-mini 0.85 Jan 2025 Official leaderboard
20 GPT-4o mini 0.83 Jul 2024 Official leaderboard
21 Amazon Nova Micro 0.83 Dec 2024 Official leaderboard
22 Qwen2.5-72B 0.82 Sep 2024 Official leaderboard
23 Mistral Large 2 Nov 2024 0.81 Nov 2024 Official leaderboard
24 GPT-4.1 nano 0.80 Apr 2025 Official leaderboard
25 Claude 3.5 Sonnet (October 2024) 0.80 Oct 2024 Official leaderboard
26 Llama-3.3-70B-Instruct-Turbo 0.77 Official leaderboard
27 Claude 3.5 Haiku 0.77 Oct 2024 Official leaderboard
28 Mistral Small v24.09 0.75 Sep 2024 Official leaderboard
29 Llama 4 Maverick 0.74 Apr 2025 Official leaderboard
30 Ministral 8B 0.69 Oct 2024 Official leaderboard
31 Meta-Llama-3.1-8B-Instruct-Turbo 0.68 Official leaderboard
32 Mistral NeMo 0.66 Jul 2024 Official leaderboard
33 Llama 4 Scout 0.63 Apr 2025 Official leaderboard

3 models have held number one since its release in Feb 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI Google DeepMind Anthropic
Own leaderboard Replaced by v2, Jul 2025
Claude 3.7 Sonnet 0.95

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
0.60 0.70 0.80 0.90 1.00 2025 0.60 0.70 0.80 0.90 1.00 2025

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

Versions

Replaced by Galileo Agent Leaderboard v2

More in Tool use

All 125 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Carbon 53.9%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Muse Spark 1.1 88.1%
  • BFCL V4 In use Models call functions and tools, including multi-turn agentic tasks with web search and memory. Checked against expected calls and outcomes. Qwen3.7-Max 75%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Claw-Eval In use 300 human-verified agent tasks (service orchestration, multimodal, multi-turn dialogue) audited via traces and snapshots. Rubric scores with Pass^3 consistency. Claude Sonnet 4.6 81.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%