LMArena Vision: creative writing leaderboard
- Claude Fable 5 leads
- 1338
- models measured
- 70
- lead over #2
- +6 Elo
- median
- 1252
Leader on LMArena Vision: creative writing · read October 22
On October 22, 2026, 71 of our catalog's models had a LMArena Vision: creative writing result, and Anthropic Claude Fable 5 led with 1338, ahead of Anthropic Claude Opus 4.6 at 1331.
Creative writing prompted by an image. People ask a question, compare two anonymous answers and vote for the better one; the rating is built from those votes, as LMArena publishes it.
Role in the index
LMArena Vision: creative writing is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Votes | Interval (±) | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5Anthropic | 1338 | 1,159 | ±18.6 | 79.3 | $50.00 | |
| 2 | Claude Opus 4.6Anthropic | 1331 | 1,407 | ±16.8 | 57.2 | $25.00 | |
| 3 | Claude Fable 5.1Anthropic | 1330 | 511 | ±28.1 | 84.2 | $50.00 | |
| 4 | Claude Opus 4.7Anthropic | 1326 | 1,454 | ±16.3 | 67.5 | $25.00 | |
| 5 | Gemini 3.1 Pro PreviewGoogle | 1324 | 3,296 | ±11.2 | 32.3 | $12.00 | |
| 6 | Claude Opus 5.5Anthropic | 1320 | 283 | ±37.4 | 88.8 | $20.00 | |
| 7 | Claude Opus 5Anthropic | 1313 | 1,633 | ±15.7 | 83.0 | $25.00 | |
| 8 | Qwen3.8 Max (0902)Qwen | 1312 | 747 | ±22.1 | 62.4 | $6.00 | |
| 9 | Gemini 3.7 FlashGoogle | 1312 | 505 | ±27.9 | 58.0 | $3.75 | |
| 10 | GPT-5.5OpenAI | 1301 | 2,038 | ±13.9 | 71.3 | $30.00 | |
| 11 | Gemini 3.5 FlashGoogle | 1301 | 1,180 | ±18.3 | 39.8 | $9.00 | |
| 12 | Claude Opus 4.8Anthropic | 1301 | 1,398 | ±16.8 | 65.0 | $25.00 | |
| 13 | GPT-6.1 SolOpenAI | 1301 | 229 | ±39.9 | 82.4 | $10.00 | |
| 14 | Gemini 3.8 FlashGoogle | 1299 | 459 | ±28.8 | 61.6 | $3.75 | |
| 15 | Muse Spark 1.3Meta | 1296 | 493 | ±27.8 | 57.4 | $4.25 | |
| 16 | Grok 4.5xAI | 1293 | 1,094 | ±19.0 | 58.6 | $6.00 | |
| 17 | GPT-6 AstraOpenAI | 1286 | 408 | ±30.8 | 85.1 | $50.00 | |
| 18 | Gemma 4 31BGoogle | 1285 | 2,703 | ±12.0 | 39.0 | $0.40 | |
| 19 | GPT-5.6 SolOpenAI | 1283 | 1,039 | ±19.4 | 73.5 | $20.00 | |
| 20 | GPT-5.4OpenAI | 1283 | 2,112 | ±13.8 | 64.4 | $15.00 | |
| 21 | Gemini 3.6 FlashGoogle | 1278 | 792 | ±22.0 | 62.5 | $3.75 | |
| 22 | DeepSeek V4.1 FlashDeepSeek | 1275 | 345 | ±32.9 | 62.5 | $0.84 | |
| 23 | Kimi K2.6Moonshot AI | 1273 | 916 | ±20.0 | 52.1 | $4.00 | |
| 24 | Claude Sonnet 4.6Anthropic | 1272 | 1,696 | ±15.2 | 57.3 | $15.00 | |
| 25 | GLM 5.3 FlashZ.ai | 1270 | 1,355 | ±16.9 | 57.7 | $0.50 | |
| 26 | Kimi K2.5Moonshot AI | 1269 | 1,768 | ±14.4 | 38.6 | $2.85 | |
| 27 | GPT-5.1OpenAI | 1266 | 935 | ±21.5 | 40.4 | $10.00 | |
| 28 | Gemini 2.5 ProGoogle | 1265 | 3,682 | ±12.4 | 19.1 | $10.00 | |
| 29 | Grok 4.6xAI | 1262 | 714 | ±23.5 | 65.7 | $6.00 | |
| 30 | Qwen3.5 397B A17BQwen | 1260 | 2,304 | ±12.9 | Not benchmarked | $2.34 | |
| 31 | Gemini 3.5 Flash LiteGoogle | 1260 | 862 | ±21.1 | 49.0 | $2.50 | |
| 32 | Qwen3.7 PlusQwen | 1258 | 1,094 | ±18.5 | 33.2 | $1.28 | |
| 33 | GPT-5.6 LunaOpenAI | 1257 | 977 | ±19.9 | 58.0 | $1.20 | |
| 34 | Gemma 4 26B A4BGoogle | 1256 | 1,495 | ±15.9 | Not benchmarked | $0.60 | |
| 35 | Muse Spark 1.1Meta | 1253 | 1,130 | ±18.8 | 70.3 | $4.25 | |
| 36 | GLM 5V TurboZ.ai | 1252 | 2,924 | ±11.7 | Not benchmarked | $4.00 | |
| 37 | GPT-5.6 TerraOpenAI | 1248 | 992 | ±19.9 | 63.9 | $12.00 | |
| 38 | Gemini 3.1 Flash Lite PreviewGoogle | 1248 | 1,883 | ±14.3 | 22.7 | $1.50 | |
| 39 | Claude Sonnet 5Anthropic | 1246 | 1,124 | ±18.8 | 57.7 | $10.00 | |
| 40 | Claude Sonnet 5.5Anthropic | 1242 | 228 | ±40.1 | 87.3 | $10.00 | |
| 41 | Gemini 2.5 FlashGoogle | 1241 | 2,919 | ±12.7 | Not benchmarked | $2.50 | |
| 42 | Qwen3.8 27BQwen | 1237 | 696 | ±23.2 | 74.2 | $2.55 | |
| 43 | Grok 4.3xAI | 1237 | 2,011 | ±14.1 | 37.4 | $2.50 | |
| 44 | MiniMax M3MiniMax | 1234 | 1,339 | ±17.0 | 39.2 | $1.20 | |
| 45 | MiMo-V2.6-FlashXiaomi | 1231 | 303 | ±35.8 | 54.7 | $0.28 | |
| 46 | MiMo-V2.5Xiaomi | 1230 | 1,908 | ±14.4 | 50.0 | $0.336 | |
| 47 | GPT-5.2OpenAI | 1225 | 1,290 | ±17.3 | 55.9 | $14.00 | |
| 48 | Qwen3.5-27BQwen | 1224 | 2,140 | ±13.6 | Not benchmarked | $1.56 | |
| 49 | GPT-5.4 MiniOpenAI | 1222 | 1,629 | ±15.5 | 35.8 | $4.50 | |
| 50 | Qwen3.5-122B-A10BQwen | 1220 | 865 | ±20.7 | Not benchmarked | $2.08 | |
| 51 | Qwen3 VL 235B A22B InstructQwen | 1220 | 974 | ±19.9 | Not benchmarked | $1.90 | |
| 52 | GPT-6 SolOpenAI | 1219 | 286 | ±37.7 | 76.0 | $10.00 | |
| 53 | Grok 4.7xAI | 1217 | 370 | ±32.5 | 66.7 | $6.00 | |
| 54 | GPT-5OpenAI | 1210 | 1,297 | ±19.4 | 38.8 | $10.00 | |
| 55 | GPT-4.1OpenAI | 1210 | 1,283 | ±19.7 | Not benchmarked | $8.00 | |
| 56 | GPT-6 LunaOpenAI | 1205 | 327 | ±34.3 | 57.3 | $0.50 | |
| 57 | T Inkling SmallThinking Machines | 1205 | 675 | ±24.0 | 27.3 | $1.20 | |
| 58 | o3OpenAI | 1201 | 1,584 | ±18.1 | Not benchmarked | $8.00 | |
| 59 | Mistral Medium 3.1Mistral AI | 1200 | 1,550 | ±17.3 | Not benchmarked | $2.00 | |
| 60 | Mistral Medium 3.5Mistral AI | 1198 | 343 | ±33.1 | 25.5 | $7.50 | |
| 61 | GPT-4.1 MiniOpenAI | 1182 | 1,192 | ±20.7 | Not benchmarked | $1.60 | |
| 62 | GPT-5 MiniOpenAI | 1181 | 1,017 | ±21.5 | 34.2 | $2.00 | |
| 63 | Gemma 3 27BGoogle | 1177 | 627 | ±25.3 | Not benchmarked | $0.30 | |
| 64 | Gemini 2.0 FlashRetiredGoogle | 1172 | 250 | ±39.0 | Not benchmarked | $0.40 | |
| 65 | o4 MiniOpenAI | 1166 | 1,258 | ±19.9 | Not benchmarked | $4.40 | |
| 66 | GPT-5.4 NanoOpenAI | 1159 | 1,531 | ±16.7 | 44.8 | $1.25 | |
| 67 | Mistral Medium 3Mistral AI | 1146 | 480 | ±28.2 | Not benchmarked | $2.00 | |
| 68 | Llama 4 ScoutMeta Llama | 1126 | 238 | ±41.0 | Not benchmarked | $0.50 | |
| 69 | Mistral Small 3.1 24BMistral AI | 1123 | 771 | ±24.5 | Not benchmarked | $0.555 | |
| 70 | GLM 4.6VZ.ai | 1113 | 241 | ±40.8 | Not benchmarked | $0.90 | |
| 71 | Llama 4 MaverickMeta Llama | 1099 | 214 | ±42.7 | Not benchmarked | $1.00 |
Only models in our catalog are listed, and each appears once with its best published run. 184 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.