Toolathlon-Verified leaderboard
- GLM 5.3 Flash leads
- 78.4%
- models measured
- 24
- lead over #2
- +1.9 pts
- median
- 58.0%
Leader on Toolathlon-Verified · read November 5
On November 5, 2026, 24 of our catalog's models had a Toolathlon-Verified result, and Z.ai GLM 5.3 Flash led with 78.4%, ahead of MoonshotAI Kimi K3 at 76.5%.
Toolathlon, the Tool Decathlon, tests language agents on 108 tasks that span 32 software applications and 604 tools, from Google Calendar and Notion to WooCommerce, Kubernetes and BigQuery. Each task takes about 20 interaction turns and is checked by its own evaluation script. Toolathlon-Verified is the current release, with revised tasks and evaluators. The number is Pass@1 as the authors publish it: the success rate of a single attempt.
How to read the score: The authors state that Toolathlon-Verified starts a new official score series: results of the original Toolathlon are not comparable and are not shown here.
Role in the index
Toolathlon-Verified is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GLM 5.3 FlashZ.ai | 78.4% | 57.7 | $0.50 | |
| 2 | Kimi K3Moonshot AI | 76.5% | 62.3 | $15.00 | |
| 3 | Claude Opus 4.8Anthropic | 76.2% | 62.8 | $25.00 | |
| 4 | Muse Spark 1.2Meta | 75.9% | 50.9 | $4.25 | |
| 5 | Muse Spark 1.1Meta | 75.6% | 70.3 | $4.25 | |
| 6 | DeepSeek V4 Pro 0813DeepSeek | 74.4% | 56.1 | $1.98 | |
| 7 | GPT-5.5OpenAI | 73.5% | 67.9 | $30.00 | |
| 8 | Claude Sonnet 5Anthropic | 71.6% | 54.7 | $10.00 | |
| 9 | DeepSeek V4 Flash 0731DeepSeek | 70.7% | 51.8 | $0.30 | |
| 10 | Gemini 3.5 FlashGoogle | 67.3% | 39.8 | $9.00 | |
| 11 | GLM 5.2Z.ai | 59.9% | 61.1 | $4.40 | |
| 12 | Kimi K2.6Moonshot AI | 58.0% | 52.1 | $4.00 | |
| 13 | Kimi K2.7 CodeMoonshot AI | 58.0% | 53.4 | $4.00 | |
| 14 | Gemini 3.5 Flash LiteGoogle | 57.1% | 48.9 | $2.50 | |
| 15 | Hy3Tencent | 56.8% | Not benchmarked | $0.528 | |
| 16 | DeepSeek V4 Pro 0423DeepSeek | 55.9% | 38.7 | $3.1275 | |
| 17 | T Inkling SmallThinking Machines | 54.4% | 27.3 | $1.20 | |
| 18 | DeepSeek V4 Flash 0423DeepSeek | 50.9% | Not benchmarked | $0.28 | |
| 19 | MiMo-V2.5Xiaomi | 49.1% | 50.0 | $2.00 | |
| 20 | MiniMax M2.7MiniMax | 47.5% | 36.2 | $1.20 | |
| 21 | T InklingThinking Machines | 45.5% | 31.6 | $4.05 | |
| 22 | Qwen3.5 397B A17BQwen | 40.7% | Not benchmarked | $2.34 | |
| 23 | Nemotron 3 UltraNVIDIA | 34.3% | 30.0 | $2.40 | |
| 24 | Kimi K2.5Moonshot AI | 33.0% | 38.6 | $2.85 |
Only models in our catalog are listed, and each appears once with its best published run. 24 rows were read from the source on November 5, 2026, and rows that match no catalog model are left out.
The board above is Toolathlon (toolathlon.xyz)'s own run. Model makers also report Toolathlon-Verified in their release announcements; those numbers are below, with the date and the announcement. They are not independent and do not count toward any index.
Other reported runs
| Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|
| DeepSeek V4 Flash 0731 | 49.7 | Preview | DeepSeek | July 31, 2026 (model release) |
| DeepSeek V4 Pro 0813 | 55.9 | Preview | DeepSeek | July 31, 2026 (model release) |