1stTop pick
Researched 10d agoClaude Opus 4.7
- Arena
- 1503
- Output (from)
- $25.00 / 1M
Best LLMs for Writing (2026)
Last refreshed 2026-06-04. Next refresh: weekly.
Top language models for long-form writing, essays, and creative prose. Ranked by Chatbot Arena human-preference scores with MMLU as a fallback.
Looking for ad copy, email, social posts, localization, or brand voice? Use the marketing use-case matrix instead of treating general prose quality as the whole decision.
Verdict
Use Claude Opus 4.7 for long-form writing today.
Claude Opus 4.6 is the runner-up: 1503 vs 1501 on Arena.
2ndShortlist
Researched 16d agoClaude Opus 4.6
- Arena
- 1501
- Output (from)
- $25.00 / 1M
3rdShortlist
Researched 1d agoGPT-5.5
- Arena
- 1488
- Output (from)
- $30.00 / 1M
How we rank
Writing picks are for essays, drafts, and long-form prose. The target rubric is EQ-Bench Creative Writing v3 when coverage is available; the live fallback is Chatbot Arena, then MMLU.
- Eligibility — General chat models excluding pure code/embedding SKUs.
- Target ranking — EQ-Bench Creative Writing v3 becomes the primary signal once it exists in seed data for enough current models.
- Current live ranking — Until that coverage lands, the page keeps the current Chatbot Arena preference score fallback, then MMLU, then newer release.
- Variant collapse — We keep one row per model family (`familySlug` + parameter tier). When headline scores tie within ±0.5 pt (±10 Elo on Chatbot Arena), we pick the canonical SKU by lowest tracked input price, then GA over preview or limited access, then newest `release`. A folded sibling within the benchmark noise band can show a "Tied within margin" chip on that score cell.
- Why this ranking? — Arena is an imperfect but useful human-preference proxy for prose quality; MMLU is only a general capability floor, not a style or brand-voice metric.
- Marketing boundary — Use `/best/marketing` for ad copy, email, social posts, localization, and brand-voice content; this page stays focused on general writing.
| # | Model | Input $/1M | Output $/1M | |
|---|---|---|---|---|
| 1 | Claude Opus 4.7 ReasoningVisionTools Arena: 1503 | $5.00 | $25.00 | |
| 2 | Claude Opus 4.6 ReasoningVisionTools Arena: 1501 | $5.00 | $25.00 | |
| 3 | Gemini 3.1 Pro Preview PreviewVisionTools Arena: 1493 | $2.00 | $12.00 | |
| 4 | Muse Spark ReasoningVisionTools Arena: 1491 | — | — | |
| 5 | GPT-5.5 ReasoningVisionTools Arena: 1488 | $5.00 | $30.00 | |
| 6 | Gemini 3 Pro VisionTools Arena: 1486 | $1.25 | $5.00 | |
| 7 | GPT-5.4 ReasoningTools Arena: 1479 | $2.50 | $15.00 | |
| 8 | ERNIE 5.1 Tools Arena: 1476 | $0.59 | $2.65 | |
| 9 | Qwen3.7-Max ReasoningTools Arena: 1475 | $1.25 | $3.75 | |
| 10 | GLM-5.1 ReasoningTools Arena: 1472 | $0.98 | $3.08 | |
| 11 | Gemini 3 Flash PreviewVisionTools Arena: 1467 | $0.50 | $3.00 | |
| 12 | Claude Opus 4.5 ReasoningVisionTools Arena: 1466 | $5.00 | $25.00 | |
| 13 | Grok 4.1 ReasoningTools Arena: 1464 | — | — | |
| 14 | DeepSeek V4 Pro ReasoningTools Arena: 1460 | $0.43 | $0.87 | |
| 15 | Claude Sonnet 4.6 ReasoningVisionTools Arena: 1459 | $3.00 | $15.00 | |
| 16 | Gemini 3.1 Flash-Lite VisionTools Arena: 1432 | $0.25 | $1.50 | |
| 17 | o3 ReasoningVisionTools Arena: 1412 | $2.00 | $8.00 | |
| 18 | Gemini 2.5 Pro ReasoningVisionTools Arena: 1398 | $1.25 | $10.00 | |
| 19 | DeepSeek R1 Reasoning Arena: 1372 | $0.10 | $0.30 | |
| 20 | Llama 4 Maverick 17B Instruct Arena: 1365 | $0.24 | $0.97 |
Honorable mentions
Next seats in this ranking. Lines below are from each model's stored description in LLMReference seed data—spot-check the model page before relying on a capability claim.
Google DeepMind's most advanced reasoning Gemini model. Part of the Gemini 3 series with frontier-class intelligence, multimodal understanding, and 1M token context window.
1486
Arena
- #5GPT-5.4
GPT-5.4 is OpenAI's flagship frontier reasoning model, released March 5, 2026. It incorporates advances from GPT-5.3-Codex for coding and agentic workflows, and adds 'Thinking' mode with editable reasoning plans. Key capabilities include computer use (navigating interfaces via Playwright), image understanding and generation integration, full-stack web app generation, tool calling, and deep research. Knowledge cutoff is August 31, 2025. Model ID: gpt-5.4.
1479
Arena
ERNIE 5.1 is Baidu's fifth-generation flagship language model, officially released May 9, 2026. Achieved via disaggregated fully-asynchronous reinforcement learning and scaled agentic post-training, it delivers leading performance at approximately 6% of the pre-training compute cost of comparable models — with roughly one-third the total parameters and half the active parameters of ERNIE 5.0. ERNIE 5.1 ranks #4 globally and #1 among Chinese models on the LMArena Search leaderboard (score: 1,223), with standout performance in legal reasoning, mathematics (AIME26: 99.6), and business domains. API model ID: ernie-5.1. Context: 128K tokens; max output: 65,536 tokens.
1476
Arena
Compare Top Picks
Side-by-side comparison of the top picks by price, benchmark, and API access.
Browse Other Categories
Best LLMs for Code GenerationBest LLMs for RAGBest AI Agents & Agentic ModelsBest LLMs for ClassificationBest Open Source LLMsBest Multimodal / Vision LLMsBest LLMs for Reasoning & MathBest Small Language Models (SLMs)Best LLMs for Function Calling & Tool UseCheapest LLM APIs You Can Call Right NowBest Long Context LLMsBest Mainstream LLM APIs, RankedBest LLMs for EnterpriseBest Free LLMs You Can Use Right NowBest LLMs for MarketingBest LLMs for Customer Support