TaxEval (Vals AI) leaderboard
- Muse Spark 1.2 leads
- 80.4%
- models measured
- 92
- lead over #2
- +0.7 pts
- median
- 73.1%
Leader on TaxEval (Vals AI) · read October 22
On October 22, 2026, 97 of our catalog's models had a TaxEval (Vals AI) result, and Meta Muse Spark 1.2 led with 80.4%, ahead of Meta Muse Spark 1.1 at 79.7%.
Tax questions from Vals AI's own dataset. Share answered correctly. An archived board.
Role in the index
TaxEval (Vals AI) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Latency, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | Muse Spark 1.2Meta | 80.4% | 48 s | 55.7 | $4.25 | |
| 2 | Muse Spark 1.1Meta | 79.7% | 33 s | 70.3 | $4.25 | |
| 3 | Claude Sonnet 4.6Anthropic | 77.1% | 2m 08s | 57.3 | $15.00 | |
| 4 | Claude Fable 5Anthropic | 76.9% | 57 s | 79.3 | $50.00 | |
| 5 | GPT-5.6 LunaOpenAI | 76.2% | 1m 17s | 58.0 | $1.20 | |
| 6 | GPT-5.6 TerraOpenAI | 76.2% | 37 s | 63.9 | $12.00 | |
| 7 | Claude Fable 5.1Anthropic | 76.0% | 1m 26s | 84.2 | $50.00 | |
| 8 | Claude Opus 4.6Anthropic | 76.0% | 1m 19s | 57.2 | $25.00 | |
| 9 | Grok 3RetiredxAI | 75.9% | 13 s | Not benchmarked | $15.00 | |
| 10 | GPT-5.2OpenAI | 75.8% | 1m 00s | 55.9 | $14.00 | |
| 11 | Kimi K3Moonshot AI | 75.7% | 2m 25s | 62.3 | $15.00 | |
| 12 | Claude Opus 4.8Anthropic | 75.6% | 1m 10s | 65.0 | $25.00 | |
| 13 | Claude Sonnet 5Anthropic | 75.6% | 3m 22s | 57.7 | $10.00 | |
| 14 | GLM 5.3 FlashZ.ai | 75.6% | 53 s | 57.7 | $0.50 | |
| 15 | Qwen3.8 Max (0902)Qwen | 75.5% | 4m 26s | 62.4 | $6.00 | |
| 16 | T Inkling SmallThinking Machines | 75.5% | 4m 02s | 27.3 | $1.20 | |
| 17 | Qwen3.7 MaxQwen | 75.3% | 58 s | 54.1 | $4.425 | |
| 18 | T InklingThinking Machines | 75.3% | 3m 18s | 31.5 | $4.05 | |
| 19 | Claude Opus 4.7Anthropic | 75.3% | 1m 02s | 67.5 | $25.00 | |
| 20 | GPT-5 MiniOpenAI | 75.2% | 37 s | 34.2 | $2.00 | |
| 21 | Claude Opus 5Anthropic | 75.1% | 58 s | 83.0 | $25.00 | |
| 22 | GPT-4.1OpenAI | 75.1% | 7.9 s | Not benchmarked | $8.00 | |
| 23 | GPT-5.5OpenAI | 75.0% | 1m 36s | 71.3 | $30.00 | |
| 24 | Claude Opus 4.5Anthropic | 74.9% | 47 s | 44.0 | $25.00 | |
| 25 | Gemini 3.6 FlashGoogle | 74.9% | 12 s | 62.5 | $3.75 | |
| 26 | GPT-5.1OpenAI | 74.9% | 44 s | 40.4 | $10.00 | |
| 27 | GPT-5.6 SolOpenAI | 74.8% | 1m 02s | 73.5 | $20.00 | |
| 28 | o4 MiniOpenAI | 74.8% | 15 s | Not benchmarked | $4.40 | |
| 29 | Gemini 3.7 FlashGoogle | 74.7% | 7.8 s | 58.0 | $3.75 | |
| 30 | Qwen3.6 PlusQwen | 74.7% | 1m 11s | 44.1 | $1.95 | |
| 31 | Kimi K2.6Moonshot AI | 74.7% | 4m 27s | 52.1 | $4.00 | |
| 32 | o3OpenAI | 74.6% | 24 s | Not benchmarked | $8.00 | |
| 33 | GPT-4o (2024-11-20)OpenAI | 74.5% | 5.9 s | Not benchmarked | $10.00 | |
| 34 | Gemini 3.8 FlashGoogle | 74.5% | 8.8 s | 61.6 | $3.75 | |
| 35 | Gemini 3.5 FlashGoogle | 74.4% | 20 s | 39.8 | $9.00 | |
| 36 | o1OpenAI | 74.3% | 19 s | Not benchmarked | $60.00 | |
| 37 | Kimi K2.5Moonshot AI | 74.2% | 1m 30s | 38.6 | $2.85 | |
| 38 | Grok 4.20xAI | 74.1% | 13 s | 33.3 | $2.50 | |
| 39 | GPT-5.4OpenAI | 74.0% | 1m 19s | 64.4 | $15.00 | |
| 40 | Gemini 3 Flash PreviewGoogle | 73.9% | 12 s | 43.7 | $3.00 | |
| 41 | MiMo-V2.5-ProXiaomi | 73.8% | 1m 03s | 31.6 | $0.9152 | |
| 42 | Claude Opus 4.1Anthropic | 73.7% | 27 s | Not benchmarked | $75.00 | |
| 43 | GPT-5OpenAI | 73.4% | 1m 19s | 38.8 | $10.00 | |
| 44 | GLM 5.2Z.ai | 73.3% | 1m 05s | 61.1 | $4.40 | |
| 45 | Claude Sonnet 4.5Anthropic | 73.3% | 48 s | 39.5 | $15.00 | |
| 46 | Nemotron 3 UltraNVIDIA | 73.1% | 36 s | 30.0 | $2.40 | |
| 47 | DeepSeek V4 Pro 0813DeepSeek | 73.1% | 3m 46s | 56.1 | $1.98 | |
| 48 | Mistral Large 3 2512Mistral AI | 73.1% | 30 s | Not benchmarked | $1.50 | |
| 49 | Gemini 3.1 Pro PreviewGoogle | 72.9% | 42 s | 32.3 | $12.00 | |
| 50 | MiniMax M3MiniMax | 72.7% | 1m 34s | 39.2 | $1.20 | |
| 51 | Gemini 3.5 Flash LiteGoogle | 72.6% | 8.8 s | 49.0 | $2.50 | |
| 52 | GLM 4.5Z.ai | 72.4% | 2m 09s | Not benchmarked | $2.20 | |
| 53 | GLM 5.3Z.ai | 72.4% | 1m 38s | 70.6 | $4.40 | |
| 54 | R1RetiredDeepSeek | 72.3% | 2m 33s | Not benchmarked | $2.50 | |
| 55 | Qwen3.5-FlashQwen | 72.2% | 1m 13s | 41.1 | $0.26 | |
| 56 | DeepSeek V4 Pro 0423DeepSeek | 72.1% | 3m 10s | 38.7 | $2.976 | |
| 57 | GPT-4.1 MiniOpenAI | 71.9% | 5.9 s | Not benchmarked | $1.60 | |
| 58 | MiMo-V2.5Xiaomi | 71.8% | 26 s | 50.0 | $0.336 | |
| 59 | Gemini 3.1 Flash Lite PreviewGoogle | 71.8% | 7.6 s | 22.7 | $1.50 | |
| 60 | Kimi K2 ThinkingMoonshot AI | 71.7% | 1m 01s | Not benchmarked | $2.50 | |
| 61 | Grok 4.5xAI | 71.7% | 3m 15s | 58.6 | $6.00 | |
| 62 | gpt-oss-120bOpenAI | 71.6% | 48 s | Not benchmarked | $0.50 | |
| 63 | Qwen3.6 27BQwen | 71.3% | 2m 28s | 36.8 | $2.70 | |
| 64 | GPT-5.4 MiniOpenAI | 71.2% | 45 s | 35.8 | $4.50 | |
| 65 | GLM 5.1Z.ai | 71.2% | 1m 24s | 49.2 | $4.40 | |
| 66 | GPT-4o (2024-08-06)OpenAI | 71.1% | 9.5 s | Not benchmarked | $10.00 | |
| 67 | DeepSeek V3 0324DeepSeek | 71.1% | 34 s | Not benchmarked | $1.07 | |
| 68 | Grok 4.6xAI | 71.1% | 48 s | 65.7 | $6.00 | |
| 69 | Qwen3.8 27BQwen | 70.8% | 1m 34s | 74.2 | $2.55 | |
| 70 | Grok 4.3xAI | 70.8% | 6m 32s | 37.4 | $2.50 | |
| 71 | DeepSeek V4 Flash 0731DeepSeek | 70.7% | 1m 12s | 51.8 | $0.396 | |
| 72 | Ling 3.0 FlashinclusionAI | 70.7% | 7.5 s | 29.6 | $0.1215 | |
| 73 | Mistral Medium 3.1Mistral AI | 70.3% | 11 s | Not benchmarked | $2.00 | |
| 74 | Kimi K2 0711Moonshot AI | 70.2% | 13 s | Not benchmarked | $2.30 | |
| 75 | GLM 5Z.ai | 70.0% | 2m 29s | 41.7 | $3.20 | |
| 76 | o3 MiniOpenAI | 69.4% | 1m 14s | Not benchmarked | $4.40 | |
| 77 | GLM 4.7Z.ai | 68.8% | 3m 07s | 55.1 | $2.20 | |
| 78 | DeepSeek V3.2DeepSeek | 68.2% | 1m 39s | 30.0 | $0.88 | |
| 79 | MiniMax M2.5MiniMax | 68.2% | 1m 13s | 37.2 | $1.20 | |
| 80 | Mistral Medium 3.5Mistral AI | 68.0% | 1m 42s | 25.5 | $7.50 | |
| 81 | Claude Haiku 4.5Anthropic | 67.5% | 48 s | 35.6 | $5.00 | |
| 82 | GPT-5.4 NanoOpenAI | 67.4% | 16 s | 44.8 | $1.25 | |
| 83 | GPT-5 NanoOpenAI | 67.4% | 1m 10s | Not benchmarked | $0.40 | |
| 84 | Llama 4 MaverickMeta Llama | 66.6% | 25 s | Not benchmarked | $1.00 | |
| 85 | MiniMax M2.7MiniMax | 66.6% | 29 s | 36.2 | $1.20 | |
| 86 | MiniMax M2.1MiniMax | 66.3% | 1m 53s | Not benchmarked | $1.20 | |
| 87 | Gemini 2.5 Flash LiteGoogle | 66.2% | 3.6 s | Not benchmarked | $0.40 | |
| 88 | GLM 4.6Z.ai | 66.2% | 1m 04s | 8.8 | $2.20 | |
| 89 | Gemini 2.0 FlashRetiredGoogle | 65.3% | 5.7 s | Not benchmarked | $0.40 | |
| 90 | Mistral LargeMistral AI | 63.8% | 12 s | Not benchmarked | $6.00 | |
| 91 | gpt-oss-20bOpenAI | 63.7% | 49 s | Not benchmarked | $0.15 | |
| 92 | Command ACohere | 61.4% | 11 s | Not benchmarked | $10.00 | |
| 93 | GPT-4.1 NanoOpenAI | 60.8% | 3.0 s | Not benchmarked | $0.40 | |
| 94 | GPT-4o-miniOpenAI | 60.5% | 8.9 s | Not benchmarked | $0.60 | |
| 95 | Gemini 1.5 ProRetiredGoogle | 59.5% | 9.9 s | Not benchmarked | $5.00 | |
| 96 | Llama 4 ScoutMeta Llama | 55.2% | 7.0 s | Not benchmarked | $0.50 | |
| 97 | Gemini 1.5 FlashRetiredGoogle | 48.2% | 3.7 s | Not benchmarked | $0.30 |
Only models in our catalog are listed, and each appears once with its best published run. 137 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.