TRAJECT-Bench
Agents select and execute tools to answer queries. The evaluation checks whether the actions satisfy a reference trajectory and produce the correct answer.
The authors publish results for ten models on parallel queries, with separate scores for simple and hard queries.
- Made by
- Pengfei He and collaborators, including Hippocratic AI
- Released
- August 2025
- Kind
- Function calling
- Level
- University
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- APIs and tools
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- GPT-5 mini #92 on the capability index
- Last model added
- September 2025
- Scores refreshed
- November 21st, 2026
Claude-4 tops the leaderboard with 71.1% Gemini 2.5 Pro leads with 70.5% among models released by September 2025
Best score per entry from the benchmark's paper, read on November 21st, 2026. A paper’s table doesn’t add newer models. The headline is the plain mean of the published Simple and Hard solution accuracy percentages. 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 best model we track is Gemini 2.5 Pro, at 70.5%. This view shows only models we track, each at its best.
The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 Claude-4 71.1%
- 2
Gemini 2.5 Pro Jun 2025 70.5%
- 3
Qwen3-235B-A22B 68.8%
- 4 Kimi K2 Sep 2025 67.5%
- 5 DeepSeek 67.3%
- 6 o4-mini 66.8%
- 7 GPT-5 mini 60.7%
- 8 gpt-oss-120b 55.6%
- 9
Gemini 2.5 Flash Sep 2025 52.3%
- 10 Claude-3.7 48%
Source: The TRAJECT-Bench paper
- 1
Gemini 2.5 Pro Jun 2025 70.5%
- 2
Qwen3-235B-A22B 68.8%
- 3 Kimi K2 Sep 2025 67.5%
- 4 o4-mini 66.8%
- 5 GPT-5 mini 60.7%
- 6 gpt-oss-120b 55.6%
- 7
Gemini 2.5 Flash Sep 2025 52.3%
Source: The TRAJECT-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude-4 | 71.1% | Official leaderboard | ||
| 2 | | 70.5% | Jun 2025 | Official leaderboard | |
| 3 | | 68.8% | Apr 2025 | Official leaderboard | |
| 4 | Kimi K2 Sep 2025 | 67.5% | Sep 2025 | Official leaderboard | |
| 5 | DeepSeek | 67.3% | Official leaderboard | ||
| 6 | o4-mini | 66.8% | Apr 2025 | Official leaderboard | |
| 7 | GPT-5 mini | 60.7% | Aug 2025 | Official leaderboard | |
| 8 | gpt-oss-120b | 55.6% | Aug 2025 | Official leaderboard | |
| 9 | | 52.3% | Sep 2025 | Official leaderboard | |
| 10 | Claude-3.7 | 48% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 70.5% | Jun 2025 | Official leaderboard | |
| 2 | | 68.8% | Apr 2025 | Official leaderboard | |
| 3 | Kimi K2 Sep 2025 | 67.5% | Sep 2025 | Official leaderboard | |
| 4 | o4-mini | 66.8% | Apr 2025 | Official leaderboard | |
| 5 | GPT-5 mini | 60.7% | Aug 2025 | Official leaderboard | |
| 6 | gpt-oss-120b | 55.6% | Aug 2025 | Official leaderboard | |
| 7 | | 52.3% | Sep 2025 | Official leaderboard |
Rankings by part
We read 2 parts from the benchmark’s paper: Simple solution accuracy and Hard solution accuracy. Each ranks models on one part of the test. Gemini 2.5 Pro leads both. Arrows show how far a model moved from its overall rank among the models we track.
Simple solution accuracy
7 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 91.1%
- 2 Kimi K2 Sep 2025 ↑1 90.2%
- 3
Qwen3-235B-A22B ↓1 89.8%
- 4 o4-mini 86.3%
- 5 GPT-5 mini 83.4%
- 6
Gemini 2.5 Flash Sep 2025 ↑1 78.2%
- 7 gpt-oss-120b ↓1 72.6%
Hard solution accuracy
7 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 49.8%
- 2
Qwen3-235B-A22B 47.9%
- 3 o4-mini ↑1 47.2%
- 4 Kimi K2 Sep 2025 ↓1 44.8%
- 5 gpt-oss-120b ↑1 38.7%
- 6 GPT-5 mini ↓1 38%
- 7
Gemini 2.5 Flash Sep 2025 26.3%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The TRAJECT-Bench paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official TRAJECT-Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More 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 Neon 56.3%
- MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks.
Gemini 4 Argon 90.7%
- 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%
- Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
- DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (10 entries).
- Pass Models scored 7 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 mini (#92).
- Pass Headroom The best is 70.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 7 rows for models we track and 3 other entries at the last read.
- Pass Last read November 21st, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.