Best LLM for Creative Writing: AI Writing Leaderboard
The writing board covers five writing-related categories of the LMArena text leaderboard, one per tab: creative writing (creative_writing, default), writing/literature/language tasks, longer prompts (longer_query; not a long-context test), multi-turn (multi_turn) and instruction following (instruction_following). Scores come from blind human pairwise votes, not automated benchmarks.
"Instruction following" is LMArena's instruction_following category: only prompts with explicit format, length or constraint requirements, reflecting how well a model follows them.
- Claude vs GeminiClaude Opus 5.5 (High)1538.8 vs 1527.6 · Δ11vs Gemini 4 Argon (High)
- Cheapest in top 20
GLM-5.3-Flash$0.238 / 1MScore 1479.1 · #20 - Best score per $ in top 20GLM-5.3-Flash6,228 / $Score 1479.1 · $0.238
Claude vs Gemini: which is better for writing?
| Model | Score | Rank | Creative writing | Input / output | Score per $ |
|---|---|---|---|---|---|
| Claude Opus 5.5 (High)Top Claude | 1538.8±10.7 | 1 | 1534.2 #1 | $4 / $20 | 192 |
| Claude Sonnet 5.5 (xHigh)Cheapest Claude in top 20 | 1490.5±11.2 | 13 | 1468.3 #18 | $2 / $10 | 373 |
 Gemini 4 Argon (High)Top Gemini | 1527.6±10.4 | 2 | 1523.1 #2 | $2 / $10 | 382 |
| Gemini 3.7 Flash (High)Cheapest Gemini in top 20 | 1486±6.8 | 15 | 1489.1 #9 | $0.75 / $3.75 | 991 |
When the gap is smaller than the combined confidence intervals, the rank difference may not be significant. Prices are OpenRouter quotes (USD / 1M tokens).
Full ranking · Instruction following
418 modelsSorted by LMArena score; ± is the 95% CI half-width. Prices: OpenRouter quote (official list price if none), USD per 1M tokens.
| # | Model | Score↑ | Rank | Votes | Creative writing | Input / output | Blended | Score per $ |
|---|---|---|---|---|---|---|---|---|
| 1 |  LLaMA 13BMeta |
780.8±24.1 | 418 | 584 | 793.7 | No price yet | ||
| 2 | ?Dolly 2.0 12B | 813.6±20.7 | 417 | 899 | 863.5 | No price yet | ||
| 3 | ?StableLM Tuned Alpha 7B | 830.9±19.8 | 416 | 814 | 867.6 | No price yet | ||
| 4 | ?FastChat-T5 3B | 858±18.2 | 415 | 1,134 | 901.9 | No price yet | ||
| 5 | ?RWKV-4 Raven 14B | 882.5±16.5 | 414 | 1,375 | 933.5 | No price yet | ||
| 6 | ?OpenAssistant Pythia 12B | 887.3±15.5 | 413 | 1,744 | 923.1 | No price yet | ||
| 7 | ?Alpaca 13B | 891.1±16.1 | 412 | 1,512 | 967.3 | No price yet | ||
| 8 | ?ChatGLM 6B | 901.8±17.4 | 411 | 1,277 | 899.8 | No price yet | ||
| 9 | ?ChatGLM2 6B | 902±22.3 | 410 | 762 | 940.9 | No price yet | ||
| 10 | ?MPT 7B Chat | 906.7±18.1 | 409 | 1,115 | 957.4 | No price yet | ||
| 11 | ?Koala 13B | 943.6±14.9 | 408 | 1,913 | 972.7 | No price yet | ||
| 12 | ?GPT4All 13B Snoozy | 945.4±24.6 | 407 | 483 | 971.8 | No price yet | ||
| 13 | ?ChatGLM3 6B | 953.9±17.6 | 406 | 1,321 | 969.8 | No price yet | ||
| 14 | Gemma 2B ITGoogle |
970.3±16 | 405 | 1,522 | 986.2 | No price yet | ||
| 15 |  OLMo 7B InstructAi2 |
978.2±15.7 | 404 | 1,908 | 989.5 | No price yet | ||
| 16 |  Qwen1.5-4B-ChatAlibaba |
978.4±12.8 | 403 | 2,636 | 969 | No price yet | ||
| 17 | ?Guanaco 33B | 991.2±20.7 | 402 | 777 | 1079.7 | No price yet | ||
| 18 | Gemma 1.1 2B ITGoogle |
991.7±10.9 | 401 | 3,877 | 997.7 | No price yet | ||
| 19 | ?Vicuna 7B | 993.7±13.9 | 400 | 2,020 | 1020.8 | No price yet | ||
| 20 |  Mistral 7B InstructMistral AI |
1003.2±13.5 | 399 | 2,768 | 1028.1 | No price yet | ||
| 21 | Llama 2 7B ChatMeta |
1005.8±9.8 | 398 | 4,541 | 1033 | No price yet | ||
| 22 | PaLM 2Google |
1010±13.3 | 397 | 2,536 | 991.2 | No price yet | ||
| 23 | ?StripedHyena Nous 7B | 1010.2±15.1 | 396 | 1,660 | 1029 | No price yet | ||
| 24 | Gemma 7B ITGoogle |
1016.9±12.7 | 395 | 2,835 | 1023.3 | No price yet | ||
| 25 |  Phi-3 Mini 128K InstructMicrosoft |
1022.7±10.1 | 394 | 7,368 | 1024 | No price yet | ||
| 26 | Code Llama 70B InstructMeta |
1023.6±29.6 | 393 | 358 | – | No price yet | ||
| 27 | Code Llama 34B InstructMeta |
1027.8±12.8 | 392 | 2,294 | 1031.4 | No price yet | ||
| 28 | ?SmolLM2 1.7B Instruct | 1029.1±21 | 391 | 859 | 1031.2 | No price yet | ||
| 29 | Llama 3.2 1B InstructMeta |
1031±11.1 | 390 | 3,248 | 1032.5 | No price yet | ||
| 30 | Qwen-14B-ChatAlibaba |
1031.2±16.4 | 389 | 1,470 | 1027.4 | No price yet | ||
| 31 | ?Zephyr 7B Beta | 1031.4±12.6 | 388 | 3,094 | 1102.1 | No price yet | ||
| 32 | ?Zephyr 7B Alpha | 1034.3±24.3 | 387 | 534 | 1068.5 | No price yet | ||
| 33 | ?Vicuna 13B | 1035±9.5 | 386 | 5,665 | 1057.9 | No price yet | ||
| 34 | Llama 2 13B ChatMeta |
1045.2±9.3 | 385 | 6,097 | 1046.6 | No price yet | ||
| 35 |  SOLAR 10.7B Instruct v1.0Upstage |
1046.4±18.1 | 384 | 1,188 | 1089.1 | No price yet | ||
| 36 | ?Falcon 180B Chat | 1046.5±28 | 383 | 389 | 1088.3 | No price yet | ||
| 37 | WizardLM 13BMicrosoft |
1048.3±13.8 | 382 | 2,003 | 1090.2 | No price yet | ||
| 38 | Phi-3 Mini 4K InstructMicrosoft |
1053.2±8.7 | 381 | 7,636 | 1036.7 | No price yet | ||
| 39 | ?Dolphin 2.2.1 Mistral 7B | 1053.3±24.3 | 380 | 497 | 1085.5 | No price yet | ||
| 40 | ?MPT 30B Chat | 1053.9±21.3 | 379 | 718 | 1073.8 | No price yet | ||
| 41 |  Granite 3.0 2B InstructIBM |
1055.6±12.2 | 378 | 2,698 | 1046 | No price yet | ||
| 42 | ?Nous Hermes 2 Mixtral 8x7B DPO | 1055.9±16.1 | 377 | 1,421 | 1101.4 | No price yet | ||
| 43 | Gemma 1.1 7B ITGoogle |
1056.8±7.9 | 376 | 8,852 | 1059.2 | No price yet | ||
| 44 | Phi-3 Mini 4K Instruct (June 2024)Microsoft |
1057.7±9.6 | 375 | 4,431 | 1044.4 | No price yet | ||
| 45 | Qwen1.5-7B-ChatAlibaba |
1058.2±13.9 | 374 | 1,715 | 1034.3 | No price yet | ||
| 46 | Mistral 7B Instruct v0.2Mistral AI |
1059.9±8.9 | 373 | 6,659 | 1067.4 | No price yet | ||
| 47 |  Llama 2 70B SteerLM ChatNVIDIA |
1060.3±17.9 | 372 | 1,076 | 1090.3 | No price yet | ||
| 48 | ?Vicuna 33B | 1064.9±8.9 | 371 | 6,983 | 1124.4 | No price yet | ||
| 49 | ?Snowflake Arctic Instruct | 1066±8.4 | 370 | 11,736 | 1080.7 | No price yet | ||
| 50 | ?Starling LM 7B Alpha | 1069.8±11.3 | 369 | 3,316 | 1098.9 | No price yet | ||
Show all 418 models (showing top 50)
Data: LMArena (lmarena-ai), CC-BY-4.0 · Instruction following · published Oct 22, 2026
Sources & method
- LMLMArena leaderboard dataset (Hugging Face, CC-BY-4.0)
config text · creative_writing · published Oct 22, 2026 · 416 models
- LMLMArena leaderboard dataset (Hugging Face, CC-BY-4.0)
config text · industry_writing_and_literature_and_language · published Oct 22, 2026 · 417 models
- LMLMArena leaderboard dataset (Hugging Face, CC-BY-4.0)
config text · longer_query · published Oct 22, 2026 · 396 models
- LMLMArena leaderboard dataset (Hugging Face, CC-BY-4.0)
config text · multi_turn · published Oct 22, 2026 · 416 models
- LMLMArena leaderboard dataset (Hugging Face, CC-BY-4.0)
config text · instruction_following · published Oct 22, 2026 · 418 models
- $
- Only the latest LMArena publish date is used; score, CI, votes and rank are copied from the dataset.
- Reasoning tiers of one model (e.g. High / Max) are separate LMArena entries and are listed separately.
- Prices are shown only when the entry maps to our LLM leaderboard; otherwise –, never estimated.
- Score ÷ blended price (3:1 input:output). A rough value indicator only.
- Synced daily at 06:35 UTC; if the source fails, the last good snapshot is kept.
FAQ
What is the best AI for writing?
In LMArena's creative writing category published Oct 22, 2026, #1 is Anthropic's Claude Opus 5.5 (High) (1534.2, 2,453 votes).
Claude or Gemini: which is better for writing?
Top creative-writing scores: Claude Claude Opus 5.5 (High) (1534.2, output $20/1M tokens); Gemini Gemini 4 Argon (High) (1523.1, output $10/1M tokens).
What is the cheapest good LLM for writing?
Among the top 20 for creative writing, the best score per dollar is MiMo-V2.6-Pro (1466.4, #20) at about $0.544/1M tokens blended; the cheapest is MiMo-V2.6-Pro ($0.544/1M, 1466.4).
Which vendors rank highest for writing?
By the creative writing category, each vendor's best model ranks: Anthropic Claude Opus 5.5 (High) (1534.2, #1), Google Gemini 4 Argon (High) (1523.1, #2), Meta Muse Wonder (Max) (1492.6, #7), Alibaba Qwen3.8-Max (1477.4, #16), Xiaomi MiMo-V2.6-Pro (1466.4, #20).
Which LLM follows instructions best?
On LMArena's Instruction following category (published Oct 22, 2026, 418 models), the top three are Claude Opus 5.5 (High) (1538.8), Gemini 4 Argon (High) (1527.6), Claude Opus 4.6 (High) (1521).
Where do the scores come from?
From LMArena: users submit writing prompts, two anonymous models answer, users vote for the better one, and scores are fitted with a Bradley-Terry model.