Fiction.LiveBench leaderboard
- GPT-5 leads
- 97.2%
- models measured
- 28
- at the top
- Tied
- median
- 63.2%
Leader on Fiction.LiveBench · read October 22
On October 22, 2026, 33 of our catalog's models had a Fiction.LiveBench result, and OpenAI GPT-5 led with 97.2%.
Questions about long stories that need the whole text held in mind; the number is the score at 16K tokens of context.
Role in the index
Fiction.LiveBench is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-5OpenAI | 97.2% | 38.8 | $10.00 | |
| 2 | o3 ProOpenAI | 97.2% | Not benchmarked | $80.00 | |
| 3 | Gemini 2.5 Pro Preview 06-05Google | 91.7% | Not benchmarked | $10.00 | |
| 4 | o3OpenAI | 88.9% | Not benchmarked | $8.00 | |
| 5 | Kimi K2.5Moonshot AI | 86.1% | 38.6 | $2.85 | |
| 6 | Claude 3.7 SonnetRetiredAnthropic | 83.3% | Not benchmarked | $15.00 | |
| 7 | DeepSeek V3.2 ExpDeepSeek | 83.3% | Not benchmarked | $0.41 | |
| 8 | o1OpenAI | 83.3% | Not benchmarked | $60.00 | |
| 9 | o4 MiniOpenAI | 77.8% | Not benchmarked | $4.40 | |
| 10 | R1 0528DeepSeek | 75.0% | Not benchmarked | $2.18 | |
| 11 | Qwen3 235B A22B Thinking 2507Qwen | 75.0% | Not benchmarked | $3.50 | |
| 12 | Qwen3 32BQwen | 74.2% | Not benchmarked | $0.40 | |
| 13 | R1RetiredDeepSeek | 69.4% | Not benchmarked | $2.50 | |
| 14 | GPT-5 MiniOpenAI | 69.4% | 34.2 | $2.00 | |
| 15 | Kimi K2 0905Moonshot AI | 66.7% | Not benchmarked | $2.50 | |
| 16 | GPT-4.1OpenAI | 63.9% | Not benchmarked | $8.00 | |
| 17 | Qwen3 14BQwen | 62.5% | Not benchmarked | $0.23 | |
| 18 | Claude Opus 4RetiredAnthropic | 61.1% | Not benchmarked | $75.00 | |
| 19 | Gemini 2.0 FlashRetiredGoogle | 61.1% | Not benchmarked | $0.40 | |
| 20 | Kimi K2 0711Moonshot AI | 61.1% | Not benchmarked | $2.30 | |
| 21 | Grok 3RetiredxAI | 58.3% | Not benchmarked | $15.00 | |
| 22 | Qwen3 Next 80B A3B InstructQwen | 55.6% | Not benchmarked | $1.10 | |
| 23 | DeepSeek V3 0324DeepSeek | 50.0% | Not benchmarked | $1.07 | |
| 24 | o3 MiniOpenAI | 50.0% | Not benchmarked | $4.40 | |
| 25 | Llama 4 MaverickMeta Llama | 46.2% | Not benchmarked | $1.00 | |
| 26 | GPT-4.1 MiniOpenAI | 44.4% | Not benchmarked | $1.60 | |
| 27 | GPT-5 NanoOpenAI | 44.4% | Not benchmarked | $0.40 | |
| 28 | gpt-oss-120bOpenAI | 44.4% | Not benchmarked | $0.50 | |
| 29 | Qwen3 30B A3BQwen | 40.6% | Not benchmarked | $0.50 | |
| 30 | Llama 4 ScoutMeta Llama | 36.0% | Not benchmarked | $0.50 | |
| 31 | Gemma 3 27BGoogle | 33.3% | Not benchmarked | $0.30 | |
| 32 | Llama 3.3 70B InstructMeta Llama | 33.3% | Not benchmarked | $0.72 | |
| 33 | GPT-4.1 NanoOpenAI | 25.0% | Not benchmarked | $0.40 |
Only models in our catalog are listed, and each appears once with its best published run. 215 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.
| Model | Result | 0 tokens | 400 tokens | 1k tokens | 2k tokens | 4k tokens | 8k tokens | 32k tokens | 60k tokens | 120k tokens | 192k tokens |
|---|---|---|---|---|---|---|---|---|---|---|---|
| OpenAI GPT-5 | 97.2% | 100.0% | 97.2% | 94.4% | 97.2% | 97.2% | 100.0% | 97.2% | 100.0% | 96.9% | 87.5% |
| OpenAI o3 Pro | 97.2% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 94.4% | 88.9% | 65.6% |
| Google Gemini 2.5 Pro Preview 06-05 | 91.7% | 100.0% | 100.0% | 100.0% | 97.2% | 94.4% | 80.6% | 91.7% | 83.3% | 87.5% | 90.6% |
| OpenAI o3 | 88.9% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 83.3% | 100.0% | 58.1% |
| MoonshotAI Kimi K2.5 | 86.1% | 100.0% | 100.0% | 100.0% | 100.0% | 94.4% | 88.9% | 88.9% | 88.9% | 78.1% | 87.5% |
| Anthropic Claude 3.7 SonnetRetired | 83.3% | 100.0% | 100.0% | 100.0% | 97.2% | 91.7% | 97.2% | 75.0% | 69.4% | 53.1% | Not published |
| DeepSeek V3.2 Exp | 83.3% | 100.0% | 100.0% | 97.2% | 97.2% | 91.7% | 94.4% | 83.3% | 71.4% | Not published | Not published |
| OpenAI o1 | 83.3% | 100.0% | 97.2% | 100.0% | 94.4% | 94.4% | 86.1% | 83.3% | 72.2% | 53.1% | Not published |
| OpenAI o4 Mini | 77.8% | 100.0% | 100.0% | 100.0% | 100.0% | 77.8% | 66.7% | 55.6% | 66.7% | 62.5% | 43.8% |
| DeepSeek R1 0528 | 75.0% | 100.0% | 91.7% | 83.3% | 82.9% | 88.9% | 86.1% | 69.4% | 58.3% | Not published | Not published |
| Qwen3 235B A22B Thinking 2507 | 75.0% | 100.0% | 100.0% | 91.7% | 86.1% | 91.7% | 97.2% | 80.6% | 77.8% | 68.8% | Not published |
| Qwen3 32B | 74.2% | 80.0% | 90.9% | 93.8% | 76.7% | 86.7% | 80.0% | Not published | Not published | Not published | Not published |
| DeepSeek R1Retired | 69.4% | 100.0% | 82.2% | 80.6% | 76.7% | 77.8% | 83.3% | 63.9% | 66.7% | 33.3% | Not published |
| OpenAI GPT-5 Mini | 69.4% | 100.0% | 97.2% | 88.9% | 80.6% | 77.8% | 66.7% | 63.9% | 61.1% | 62.5% | 59.4% |
| MoonshotAI Kimi K2 0905 | 66.7% | 87.5% | 75.0% | 72.2% | 72.2% | 66.7% | 58.3% | 63.9% | 55.6% | 40.6% | 50.0% |
| OpenAI GPT-4.1 | 63.9% | 100.0% | 91.7% | 75.0% | 69.4% | 63.9% | 55.6% | 58.3% | 52.8% | 62.5% | 56.3% |
| Qwen3 14B | 62.5% | 83.3% | 64.5% | 61.8% | 59.4% | 64.7% | 51.6% | Not published | Not published | Not published | Not published |
| Anthropic Claude Opus 4Retired | 61.1% | 100.0% | 77.8% | 77.8% | 66.7% | 66.7% | 66.7% | 63.9% | 55.6% | 37.5% | Not published |
| Google Gemini 2.0 FlashRetired | 61.1% | 100.0% | 63.9% | 58.3% | 55.6% | 47.2% | 50.0% | 50.0% | 47.2% | 62.5% | Not published |
| MoonshotAI Kimi K2 0711 | 61.1% | 87.5% | 75.0% | 69.4% | 66.7% | 58.3% | 55.6% | 50.0% | 44.4% | Not published | Not published |
| xAI Grok 3Retired | 58.3% | 75.0% | 72.2% | 63.9% | 55.6% | 55.6% | 52.8% | 55.6% | 63.9% | 58.3% | Not published |
| Qwen3 Next 80B A3B Instruct | 55.6% | 100.0% | 75.0% | 72.2% | 63.9% | 61.1% | 58.3% | 52.8% | 55.6% | 62.5% | 46.9% |
| DeepSeek V3 0324 | 50.0% | 87.5% | 61.1% | 69.4% | 52.8% | 52.8% | 52.8% | 55.6% | 55.6% | Not published | Not published |
| OpenAI o3 Mini | 50.0% | 100.0% | 63.9% | 58.3% | 47.2% | 47.2% | 50.0% | 55.6% | 44.4% | 43.8% | Not published |
| Meta Llama 4 Maverick | 46.2% | 100.0% | 56.0% | 50.0% | 52.0% | 48.0% | 48.0% | 44.0% | 32.0% | 36.4% | Not published |
| OpenAI GPT-4.1 Mini | 44.4% | 75.0% | 66.7% | 55.6% | 41.7% | 44.4% | 41.7% | 38.9% | 38.9% | 46.9% | Not published |
| OpenAI GPT-5 Nano | 44.4% | 75.0% | 58.3% | 63.9% | 50.0% | 50.0% | 52.8% | 47.2% | 36.1% | 21.9% | 34.4% |
| OpenAI gpt-oss-120b | 44.4% | 100.0% | 75.0% | 61.1% | 55.6% | 47.2% | 44.4% | 38.9% | Not published | Not published | Not published |
| Qwen3 30B A3B | 40.6% | 85.7% | 58.1% | 54.8% | 51.5% | 53.3% | 50.0% | Not published | Not published | Not published | Not published |
| Meta Llama 4 Scout | 36.0% | 62.5% | 52.0% | 50.0% | 36.0% | 32.0% | 40.0% | 16.0% | 24.0% | 27.3% | Not published |
| Google Gemma 3 27B | 33.3% | 87.5% | 44.4% | 50.0% | 41.7% | 33.3% | 38.9% | 25.0% | 30.6% | Not published | Not published |
| Meta Llama 3.3 70B Instruct | 33.3% | 75.0% | 66.7% | 69.4% | 55.6% | 41.7% | 36.1% | 33.3% | 33.3% | Not published | Not published |
| OpenAI GPT-4.1 Nano | 25.0% | 62.5% | 50.0% | 41.7% | 36.1% | 33.3% | 38.9% | 33.3% | 36.1% | 18.8% | Not published |
The parts as Epoch AI benchmarking hub publishes them, for the same run as the result.