Claude vs ChatGPT: API prices, benchmarks and which to pick
On the API, Anthropic's Claude and OpenAI's GPT models now list at the same price in every tier we pair: Claude Fable 5.1 and GPT-6 Astra at $10.00 / $50.00, Claude Sonnet 5.5 and GPT-6.1 Sol at $2.00 / $10.00, and Claude Haiku 5.5 and GPT-6 Luna at $0.10 / $0.50 per million tokens. The differences that remain are in cached input, how long a prompt can get before the price rises, and what a finished task costs once a model has used its tokens. This page lays out the list prices, what typical workloads cost, and the public benchmark results we could find for these models, each with its source and the date we checked it.
Prices checked 10 Oct 2026
Compare two different models →API prices, tier by tier
Per 1M tokens, in / out · US dollars| Tier | Claude | ChatGPT |
|---|---|---|
| Flagship | Claude Fable 5.1$10.00 / $50.00 | GPT-6 Astra$10.00 / $50.00 |
| Mid-tier | Claude Sonnet 5.5$2.00 / $10.00 | GPT-6.1 Sol$2.00 / $10.00 |
| Budget | Claude Haiku 5.5$0.10 / $0.50 | GPT-6 Luna$0.10 / $0.50 |
| Models listed, not marked legacy | 5 | 10 |
Flagship: GPT-6 Astra's rates rise to $20/$75 on prompts over 272K tokens. Fable 5.1 bills its whole 1M window at the standard rate.
Mid-tier: GPT-6.1 Sol's rates rise on prompts over 272K tokens. Sonnet 5.5 bills its whole context window at the standard rate.
Budget: The budget match holds on short prompts only: Haiku 5.5 bills a higher rate once a prompt passes 100K tokens, GPT-6 Luna once it passes 272K.
Context, batch and cache
Cached input per 1M tokens · batch discount off list price| Tier | Claude | ChatGPT |
|---|---|---|
| Flagship | Context: 1M tokensBatch: 50% offCached input: $0.25 | Context: 1.05M tokensBatch: 50% offCached input: $1.00 |
| Mid-tier | Context: 1M tokensBatch: 50% offCached input: $0.10 | Context: 1.05M tokensBatch: 50% offCached input: $0.10 |
| Budget | Context: 1M tokensBatch: 50% offCached input: $0.010 | Context: 1.05M tokensBatch: 50% offCached input: $0.010 |
What typical workloads cost
Monthly cost at list prices, no batch or caching, for four request shapes.
Flagship: Claude Fable 5.1 vs GPT-6 Astra
| Workload | Claude Fable 5.1 | GPT-6 Astra |
|---|---|---|
| Chatbot50K requests × 1.5K in / 400 out | $1,750 | $1,750 |
| Document Q&A10K requests × 12K in / 600 out | $1,500 | $1,500 |
| Coding agent5K requests × 30K in / 2K out | $2,000 | $2,000 |
| Summaries200K requests × 4K in / 300 out | $11,000 | $11,000 |
Mid-tier: Claude Sonnet 5.5 vs GPT-6.1 Sol
| Workload | Claude Sonnet 5.5 | GPT-6.1 Sol |
|---|---|---|
| Chatbot50K requests × 1.5K in / 400 out | $350.00 | $350.00 |
| Document Q&A10K requests × 12K in / 600 out | $300.00 | $300.00 |
| Coding agent5K requests × 30K in / 2K out | $400.00 | $400.00 |
| Summaries200K requests × 4K in / 300 out | $2,200 | $2,200 |
Budget: Claude Haiku 5.5 vs GPT-6 Luna
| Workload | Claude Haiku 5.5 | GPT-6 Luna |
|---|---|---|
| Chatbot50K requests × 1.5K in / 400 out | $17.50 | $17.50 |
| Document Q&A10K requests × 12K in / 600 out | $15.00 | $15.00 |
| Coding agent5K requests × 30K in / 2K out | $20.00 | $20.00 |
| Summaries200K requests × 4K in / 300 out | $110.00 | $110.00 |
Public benchmark results
Results from third-party leaderboards, quoted as published and checked on the date shown. They are not our tests, scores move as models and test harnesses are updated, and a benchmark is not your workload.
Terminal-Bench 4.0 and Vibe Code Bench are agentic coding tests; higher is better. The Intelligence Index is Artificial Analysis's combined score, with the dollar cost of its run per task in brackets. Each Intelligence Index figure is the max reasoning setting listed for that model.
Which to pick for what
It lists at the same $2.00 / $10.00 as Sonnet 5.5, but Artificial Analysis recorded far fewer dollars per task for it on its index run (see the benchmarks above). List price alone does not predict the bill: how many tokens a model spends does.
On Vals AI's Terminal-Bench 4.0 and Vibe Code Bench it scored higher than GPT-6.1 Sol, at the same list price and with a 1M-token window billed at standard rates.
At the same $10.00 / $50.00 list price it scored slightly higher than Fable 5.1 on Terminal-Bench 4.0, and Artificial Analysis recorded a lower cost per task for it (see the Intelligence Index rows above). Fable 5.1 edged it on Vibe Code Bench.
At $4.00 / $20.00 it ranks first on Vals AI's Terminal-Bench 4.0 (65.15%) and scores 90.29% on Vibe Code Bench, above both Fable 5.1 and GPT-6 Astra on each. We pair Fable 5.1 with GPT-6 Astra by price tier, but Opus 5.5 is the cheaper Claude model to test first for agentic coding. GPT-5.6 Sol is listed at $4.00 / $20.00 too, on a promotional rate that OpenAI says runs at least through 21 November 2026; we did not pair it by tier.
Cached input is $0.25 per million tokens against $1.00 for GPT-6 Astra, at the same list price. If most of your input is a reused prefix, that gap is the main price difference between them.
Same $0.10 / $0.50 as GPT-6 Luna and scored higher on both Vals AI benchmarks above. Keep prompts under 100K tokens.
It holds its $0.10 / $0.50 rate to 272K tokens, where Haiku 5.5's rate rises after 100K. Artificial Analysis also recorded a lower cost per task for it than for Haiku 5.5 (see the Intelligence Index rows above).
Next steps
Claude vs ChatGPT FAQ
Is Claude cheaper than ChatGPT?
On list price, the two are now level at every tier. Claude Haiku 5.5 matches GPT-6 Luna's $0.10/$0.50 for prompts up to 100,000 tokens, Claude Sonnet 5.5 costs exactly the same as GPT-6.1 Sol ($2/$10), Claude Opus 5.5 matches GPT-5.6 Sol's promotional $4/$20 exactly, and Claude Fable 5.1 costs the same as GPT-6 Astra. The main gap left is long prompts at the budget end: above 100,000 tokens Haiku 5.5 bills $0.50/$2.50, while GPT-6 Luna stays at $0.10/$0.50 up to 272K tokens and $0.20/$0.75 beyond.
Do both offer batch processing?
Yes — both discount batch jobs by 50% on input and output. Anthropic supports it across all active Claude models, and OpenAI's Batch API covers the GPT family including the GPT-6 and GPT-5.6 models.