GPT vs Claude API Cost: Every Current Model Compared
OpenAI’s GPT-6 and GPT-5.6 against Anthropic’s Claude Fable, Opus, Sonnet and Haiku — list prices, real-workload costs and the details that change the math.
Updated September 14, 2026
OpenAI and Anthropic both bill their APIs per million tokens, with one rate for the tokens you send (input) and a higher one for the tokens the model writes back (output). The headline numbers look similar, but tiering, caching discounts and even the way each company counts tokens can swing a monthly bill by a wide margin. Here is how the current lineups compare, using list prices from both providers’ official pricing pages.
The lineups, tier by tier
Each provider sells a ladder of models. We paired them by price position, not by benchmark claims, so you can see what you pay at each rung. Prices are USD per 1M tokens (input / output).
| Tier | OpenAI | Price | Anthropic | Price | |
|---|---|---|---|---|---|
| Top tier | GPT-6 Astra | $10.00 / $50.00 | Claude Fable 5.1 | $10.00 / $50.00 | Compare → |
| Flagship | GPT-5.6 Sol | $4.00 / $20.00 | Claude Opus 5 | $5.00 / $25.00 | Compare → |
| Balanced | GPT-5.6 Terra | $2.00 / $12.00 | Claude Sonnet 5 | $2.00 / $10.00 | Compare → |
| Budget | GPT-5.6 Luna | $0.20 / $1.20 | Claude Haiku 4.5 | $1.00 / $5.00 | Compare → |
What real workloads cost
Per-token prices are hard to reason about, so we ran three typical workloads through each model:
- Support chatbot: 100,000 replies a month, ~1,500 input tokens (system prompt + history) and ~400 output tokens each.
- Document Q&A (RAG): 50,000 questions a month with ~8,000 tokens of retrieved context and ~600 output tokens each.
- Coding agent: 10,000 agent steps a month, ~40,000 input tokens each (70% read from the prompt cache) and ~2,000 output tokens.
| Model | Support chatbot | Document Q&A (RAG) | Coding agent |
|---|---|---|---|
| GPT-6 Astra | $3,500 | $5,500 | $2,480 |
| Claude Fable 5.1 | $3,500 | $5,500 | $2,270 |
| GPT-5.6 Sol | $1,400 | $2,200 | $992.00 |
| Claude Opus 5 | $1,750 | $2,750 | $1,240 |
| GPT-5.6 Terra | $780.00 | $1,160 | $536.00 |
| Claude Sonnet 5 | $700.00 | $1,100 | $496.00 |
| GPT-5.6 Luna | $78.00 | $116.00 | $53.60 |
| Claude Haiku 4.5 | $350.00 | $550.00 | $248.00 |
Details that change the math
Claude’s newer tokenizer counts more tokens
Anthropic notes that Claude Opus 4.7 and later models use a newer tokenizer that produces roughly 30% more tokens for the same text. That applies to Claude Fable 5.1, Opus 5 and Sonnet 5; Haiku 4.5 still uses the previous tokenizer. A price that looks identical per token can therefore mean a larger bill for the same prompt. The only reliable comparison is to send a sample of your real prompts to both APIs and read the usage numbers they return.
Prompt caching
Both providers discount repeated prompt prefixes. On OpenAI’s GPT-5.6 models and GPT-6 Astra, cached input costs 10% of the input rate, and writing to the cache costs 1.25x the input rate. Anthropic charges 10% of the base input price for cache hits, dropping to 2.5% on Claude Fable 5.1. Its cache writes cost 1.25x for a five-minute cache or 2x for a one-hour cache. For agents that resend a large context on every step, caching often matters more than the list price.
Long prompts
OpenAI lists separate long-context rates. GPT-6 Astra, for example, moves from $10.00 / $50.00 to $20.00 / $75.00 per 1M tokens on long-context requests. All four GPT models have 1.05M-token context windows. Claude Fable 5.1, Opus 5 and Sonnet 5 offer 1M tokens; Haiku 4.5 offers 200K.
Batch jobs and promotions
Both providers take 50% off work sent through their batch APIs. Two prices deserve a watch: OpenAI says GPT-5.6 Sol’s promotional pricing runs at least through November 21, 2026, and Anthropic has made Claude Sonnet 5’s $2 / $10 launch price permanent instead of raising it to $3 / $15 on September 1, 2026.
So which is cheaper?
For the document Q&A workload, tier by tier:
- Top tier: GPT-6 Astra at $5,500/month vs $5,500 for Claude Fable 5.1.
- Flagship: GPT-5.6 Sol at $2,200/month vs $2,750 for Claude Opus 5 (20% less).
- Balanced: Claude Sonnet 5 at $1,100/month vs $1,160 for GPT-5.6 Terra (5% less).
- Budget: GPT-5.6 Luna at $116.00/month vs $550.00 for Claude Haiku 4.5 (79% less).
Per-token cost is only half the equation. A cheaper model that needs more retries, longer outputs or extra guardrails can end up costing more. Plug your own token counts into the cost calculator before you decide.