The Arena leaderboard — one of the most popular blind-evaluation benchmarks in AI, where human voters judge outputs head-to-head without knowing which model produced them — has a dedicated creative writing track. The results for June 2026 tell a clear story: the best AI models for creative writing are overwhelmingly proprietary, dominated by Anthropic and Google, with xAI sneaking into the top ten. Here’s the full breakdown.
1. claude-opus-4-6-thinking
Score: 1497 | Votes: 5,508 | Price: $5 / $25 per million tokens
At the top of the leaderboard sits Anthropic’s claude-opus-4-6-thinking — and it’s earned its place. With a score of 1497 and over 5,500 human votes validating its output, this is the best AI model for creative writing available today. The “thinking” variant applies extended reasoning to prose, letting the model work through narrative structure, voice, and tone before it writes — and the results show. Claude Opus 4.6 topped the Artificial Analysis Intelligence Index on its February 2026 launch and has since demonstrated a consistent edge in tasks requiring creative nuance. At $5/$25 per million tokens, it sits at a premium price, but for anyone serious about creative output quality, the investment makes sense.
2. gemini-3-pro
Score: 1485 | Votes: 6,290 | Price: $2 / $12 per million tokens
Google’s gemini-3-pro ranks second with 1485 points and the highest vote count in the top five — over 6,290 human evaluations, giving its score strong statistical weight. As one of the best AI models for creative writing from Google, Gemini 3 Pro brings a breadth of style and a fluid, adaptable voice to everything from fiction to long-form narrative. When Gemini 3 topped all major LMArena leaderboards in late 2025, it marked a turning point for Google’s creative capabilities — and the creative writing Arena rankings confirm that momentum has held. At $2/$12 per million tokens, it’s significantly more affordable than Claude Opus, making it one of the better value picks among the best AI models for creative writing.
3. claude-opus-4-7-thinking
Score: 1485 | Votes: 3,245 | Price: $5 / $25 per million tokens
Anthropic’s claude-opus-4-7-thinking ties Gemini 3 Pro on score at 1485 but has fewer votes — 3,245 — reflecting its more recent arrival on the leaderboard. It’s a meaningful upgrade from its predecessor: Claude Opus 4.7 improved on Opus 4.6 across long-running agentic tasks, vision, and instruction-following, and notably sees images at more than three times the resolution, which translates to stronger outputs when creative work involves visual reference or multimodal prompts. Among the best AI models for creative writing, the 4.7 thinking variant is one to watch as its vote count matures. Priced identically to its predecessor at $5/$25 per million tokens, it gives Anthropic users a straightforward upgrade path.
4. claude-opus-4-7
Score: 1484 | Votes: 3,398 | Price: $5 / $25 per million tokens
Just one point behind its thinking-mode sibling sits the standard claude-opus-4-7, with 1484 points across 3,398 votes. The proximity of these two scores is telling: for creative writing, the extended reasoning of the thinking variant adds only a marginal edge in blind human evaluation. The standard model is faster and equally well-positioned among the best AI models for creative writing for most use cases — fiction, copywriting, screenwriting, and long-form prose. Anthropic’s consistent hold on four of the top ten spots on this leaderboard underscores how seriously the company has invested in prose quality as a competitive differentiator.
5. gemini-3.1-pro-preview
Score: 1483 | Votes: 7,005 | Price: $2 / $12 per million tokens
gemini-3.1-pro-preview carries the highest vote count of any model in the top ten — 7,005 — which makes its score of 1483 among the most statistically robust on the entire leaderboard. Google describes Gemini 3.1 Pro as a step forward in core reasoning with particular strength in visualizing difficult concepts and bringing creative projects to life. As one of the best AI models for creative writing from any lab, it excels at synthesizing ideas into coherent, stylistically varied prose. Its $2/$12 pricing makes it one of the most accessible high-performance options available to writers and developers right now.
6. claude-opus-4-6
Score: 1477 | Votes: 5,665 | Price: $5 / $25 per million tokens
The non-thinking version of claude-opus-4-6 sits at rank six with 1477 points — still firmly in the conversation for the best AI models for creative writing, even after being surpassed by its own successor variants. Its 5,665 votes give it a reliable signal, and its score reflects the strong baseline that Anthropic built with the 4.6 generation. Writers working with Claude Opus 4.6 via the API are getting a model that remains competitive at the top of the global creative writing benchmark, even as newer releases pile up above it.
7. claude-opus-4-5-20251101-thinking
Score: 1468 | Votes: 5,546 | Price: $5 / $25 per million tokens
The claude-opus-4-5-20251101-thinking variant — Anthropic’s dated-release thinking model from late 2025 — still ranks seventh among the best AI models for creative writing with 1468 points and over 5,500 votes. This is a model that Anthropic said outperformed all human candidates on a performance engineering exam within the standard two-hour time limit — a proxy for how it handles complex, time-pressured creative tasks as well. The thinking-mode architecture gives it a structural advantage in longer-form writing where narrative consistency and planning matter. At $5/$25, it’s the same price as the newer Claude Opus variants.
8. gemini-3.5-flash
Score: 1464 | Votes: 1,376 | Price: $1.50 / $9 per million tokens
gemini-3.5-flash is the wildcard in this list. Flagged as Preliminary — meaning it has fewer votes than the leaderboard threshold for a confirmed ranking — its score of 1464 is striking for a Flash-class model. Flash variants are designed for speed and cost-efficiency, and at $1.50/$9 per million tokens it’s the second cheapest option in the top ten. If its ranking holds as votes accumulate, it will cement itself as one of the best AI models for creative writing for high-volume use cases where cost-per-word matters. Google has a history of closing the gap between its Flash and Pro models faster than the industry expects — watch this one.
9. claude-opus-4-5-20251101
Score: 1463 | Votes: 10,245 | Price: $5 / $25 per million tokens
With 10,245 votes, claude-opus-4-5-20251101 has been evaluated by more humans than any other model in the top ten — making its score of 1463 the most statistically reliable figure on the entire list. For writers evaluating the best AI models for creative writing with confidence, this is the benchmark. The non-thinking version of Claude Opus 4.5 shows that Anthropic’s base models — without extended reasoning — are still top-tier for prose. Claude Opus 4.5 beat GPT-5.1 on the Artificial Analysis Intelligence Index and performed strongly across creative and reasoning tasks alike.
10. grok-4.20-beta1
Score: 1462 | Votes: 3,685 | Price: N/A
The only non-Anthropic, non-Google model in the top ten is xAI’s grok-4.20-beta1, scoring 1462 across 3,685 votes. Pricing is listed as N/A — it’s currently in beta. xAI has been pushing Grok into leaderboard positions across multiple benchmarks, and its appearance here among the best AI models for creative writing is a genuine signal that the model has creative chops alongside its technical capabilities. Grok also leads all major AI platforms in average session duration — suggesting users find its outputs engaging enough to keep conversations going. Whether xAI can maintain this position as the model moves out of beta and pricing is announced will be one of the more interesting stories to follow.
The Takeaway
The June 2026 Arena creative writing leaderboard is a near-clean sweep for Anthropic and Google. Anthropic holds six of the top ten spots; Google holds three. The best AI models for creative writing right now are Claude Opus (in both standard and thinking variants) and Gemini 3/3.1 Pro — with Google offering significantly better value per token. For budget-conscious users, gemini-3-pro at $2/$12 and gemini-3.5-flash at $1.50/$9 are the clear standouts. For those who want the absolute ceiling, claude-opus-4-6-thinking at the top of the leaderboard remains the answer.