Atlas

Benchmarks

← All benchmarks

Tax Agent Bench — Current & Temporal All-Pass

Professional Work · 2026-09-23

All-rubric-item pass rate on the 30 current-and-temporal questions.

Top models (higher is better)

ModelScore
Gemini 4 Argon66.7
Claude Fable 5.160.0
Muse Spark 1.256.7
Claude Fable 553.3
GLM-5.353.3
MiMo-V2.6-Pro53.3
Claude Opus 550.0
Ember-150.0
GPT-5.6 Sol50.0
Muse Spark 1.350.0
Opus 4.846.7
DeepSeek V4 Pro 081346.7
Grok 4.746.7
Kimi K346.7
Mistral Large 4 Preview46.7
DeepSeek-V4-Flash-073143.3
Gemini 3.8 Flash43.3
GLM-5.3 Flash43.3
Grok 4.643.3
Qwen3.8 27B43.3
DeepSeek V4.1 Flash40.0
Opus 4.736.7
Grok 4.536.7
Hy4 Preview36.7
MiMo-V2.6-Flash36.7
Sonnet 4.633.3
GPT-5.6 Luna33.3
GPT-5.6 Terra33.3
GPT-6.1 Sol33.3
Inkling-Small33.3
Step 5 Preview33.3
Sonnet 530.0
Gemini 3.7 Flash30.0
GPT-6 Astra30.0
MiniMax M330.0
Gemini 3.5 Flash26.7
GPT-5.526.7
Kimi K2.626.7
GPT-6 Sol23.3
Qwen3.7-Max (source unspecified)23.3
Loading Atlas data…