Benchmarks · agentic
Tax Agent Bench
Evaluating agents on research-grade US tax questions using the Tax Agent Bench harness
60 models · 61 results · updated 2026-11-05 · vals.ai ↗
Leaderboard
Gemini 4 Carbon high
Heavy line: best score at each price. Thin lines join one model's effort levels. Cost uses each model's median subscription plan, or API price with no plan data. 45 more models under All models in settings.
More from this benchmark
1 published measuresTime per task82 results · lower is better
42 more models under All models in settings.
All results
61 rows| # | Model | Effort | Tax Agent Bench | Subscription $ / task | API $ / task | Time |
|---|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon | high | 51.8% | $2.26 | $6.34 | 35 min |
| 2 | Fable 5.1 | max | 49.2% | $2.05 | $13 | 29 min |
| 3 | Opus 5.5 | max | 45.6% | $0.23 | $15 | 1.1 h |
| 4 | Opus 5 | max | 45.5% | $0.089 | $5.14 | 17 min |
| 5 | Gemini 4 Argon | high | 44.6% | $2.48 | $3.47 | 16 min |
| 6 | DeepSeek V4.1 Flash 1017 | max | 44.4% | free | $0.46 | 21 min |
| 7 | Muse Spark 1.3 | max | 42.4% | $0.19 | $0.39 | 4.1 min |
| 8 | Sonnet 5.5 | max | 42.0% | $0.14 | $9.25 | 59 min |
| 9 | GLM 5.3 | max | 39.7% | $0.19 | $1.80 | 51 min |
| 10 | Inkling 2 Small | max | 38.9% | $0.14 | $0.14 | 6.9 min |
| 11 | Fable 5 | max | 38.3% | $0.75 | $6.70 | 14 min |
| 12 | Muse Spark 1.3 | xhigh | 36.0% | $0.16 | $0.32 | 8.4 min |
| 13 | Hy4 | high | 34.4% | $0.48 | $1.69 | 1.1 h |
| 14 | Grok 4.7 | xhigh | 34.2% | $0.17 | $1.80 | 17 min |
| 15 | Kimi K3 | max | 33.4% | $0.25 | $2.79 | 40 min |
| 16 | Grok 4.6 | high | 33.1% | $0.10 | $0.98 | 11 min |
| 17 | Muse Spark 1.2 | xhigh | 32.8% | $0.13 | $0.27 | 2.0 min |
| 18 | Hy4 preview | high | 32.6% | $0.44 | $0.59 | 41 min |
| 19 | MiMo V2.6 Pro | max | 32.4% | $0.11 | $0.13 | 26 min |
| 20 | Gemini 3.8 Flash | high | 32.1% | $0.035 | $0.86 | 3.0 min |
| 21 | Qwen 3.8 Max | max | 32.1% | $0.18 | $1.55 | 41 min |
| 22 | Haiku 5.5 | max | 31.9% | $0.025 | $1.16 | 50 min |
| 23 | GLM 5.3 Flash | max | 31.6% | free | $0.14 | 22 min |
| 24 | GPT-5.6 Sol | max | 31.1% | $0.21 | $7.51 | 45 min |
| 25 | Qwen 3.8 27B | xhigh | 30.9% | $0.060 | $0.82 | 22 min |
| 26 | GPT-5.6 Terra | max | 30.0% | $0.13 | $4.83 | 1.0 h |
| 27 | Opus 4.8 | max | 30.0% | $0.029 | $1.69 | 8.0 min |
| 28 | Sonnet 5 | max | 29.4% | $0.031 | $1.79 | 19 min |
| 29 | MiMo V2.6 Flash | high | 28.5% | free | $0.049 | 17 min |
| 30 | DeepSeek V4 Flash 0731 | high | 28.1% | $0.16 | $0.16 | 5.6 min |