Home › Compare › GPT-5.6 Luna vs Claude Haiku 4.5
GPT-5.6 Luna vs Claude Haiku 4.5
A straight cost comparison of GPT-5.6 Luna (OpenAI) and Claude Haiku 4.5 (Anthropic), per 1,000,000 tokens, verified October 2026.
Rate-by-rate comparison
| Metric | GPT-5.6 Luna | Claude Haiku 4.5 | Winner |
|---|---|---|---|
| Input / 1M | $0.20 | $1 | GPT-5.6 Luna |
| Cached input / 1M | $0.02 | $0.10 | GPT-5.6 Luna |
| Cache write / 1M | $0.25 | $1.25 | GPT-5.6 Luna |
| Output / 1M | $1.20 | $5 | GPT-5.6 Luna |
| Context window | 1,050,000 | 200,000 | GPT-5.6 Luna |
| Batch API | Yes | Yes | Tie |
| Long-context tier | > 272,000 | None | — |
For per-token metrics, the lower number wins. For context window and batch support, the higher or broader option wins. A dash means the provider does not publish that rate: a missing cache-write fee means writes are not billed per token, while a missing cached-input rate means cache hits bill at the full input rate.
Blended cost: 1M input + 250K output
| Model | Blended cost | Difference |
|---|---|---|
| GPT-5.6 Luna | $0.50 | Cheaper |
| Claude Haiku 4.5 | $2.25 | +$1.75 |
Ten workloads, two bills
The same two models, priced across ten workloads that differ only in their input:output ratio. Nothing else changes — no retries, no system prompt, no middleware.
| Workload | GPT-5.6 Luna | Claude Haiku 4.5 | Winner |
|---|---|---|---|
| Classification 100K reqs · 500 in / 10 out | $11.20 | $55 | GPT-5.6 Luna |
| Data extraction 10K reqs · 2K in / 200 out | $6.40 | $30 | GPT-5.6 Luna |
| Support chat 1K reqs · 4K in / 600 out | $1.52 | $7 | GPT-5.6 Luna |
| Code review 2K reqs · 8K in / 1.5K out | $6.80 | $31 | GPT-5.6 Luna |
| Summarisation 1K reqs · 20K in / 1K out | $5.20 | $25 | GPT-5.6 Luna |
| Agent loop 5K reqs · 12K in / 2K out | $24 | $110 | GPT-5.6 Luna |
| Translation 500 reqs · 15K in / 15K out | $10.50 | $45 | GPT-5.6 Luna |
| Long-document QA 200 reqs · 120K in / 800 out | $4.99 | $24.80 | GPT-5.6 Luna |
| Cached RAG 1K reqs · 50K in (45K cached) / 1K out | $3.11 | $14.55 | GPT-5.6 Luna |
| Nightly batch 10K reqs · 10K in / 2K out | $22 | $100 | GPT-5.6 Luna |
GPT-5.6 Luna wins 10 of 10, Claude Haiku 4.5 wins 0 of 10. The winner is set by the token mix, not by the headline rate.
Where the two bills cross over
GPT-5.6 Luna is cheaper on both sides of the mix: $0.20 vs $1 per million input tokens, and $1.20 vs $5 per million output. No input:output ratio reverses that, so on price alone the other model never becomes the better choice — it has to win on context, latency or output quality.
Cost at four scales
The same support-chat workload (4K input, 600 output per request) at four monthly volumes.
| Requests / month | GPT-5.6 Luna | Claude Haiku 4.5 | Difference |
|---|---|---|---|
| 100 | $0.152 | $0.70 | GPT-5.6 Luna cheaper by $0.548 |
| 1,000 | $1.52 | $7 | GPT-5.6 Luna cheaper by $5.48 |
| 10,000 | $15.20 | $70 | GPT-5.6 Luna cheaper by $54.80 |
| 100,000 | $152 | $700 | GPT-5.6 Luna cheaper by $548 |
What caching does to each bill
Caching only helps the part of the prompt that repeats. This is the support-chat workload (1,000 requests, 4,000 input tokens each) with an increasing share of that prompt served from a warm cache, written once on the first request.
| Cached share of the prompt | GPT-5.6 Luna | Claude Haiku 4.5 | Winner |
|---|---|---|---|
| 0% | $1.52 | $7 | GPT-5.6 Luna |
| 50% | $1.16 | $5.20 | GPT-5.6 Luna |
| 80% | $0.945 | $4.12 | GPT-5.6 Luna |
| 95% | $0.837 | $3.58 | GPT-5.6 Luna |
Both models publish a cached-read rate, so the discount is comparable: GPT-5.6 Luna reads cache at $0.02 per million, Claude Haiku 4.5 at $0.10.
Which one should you pick?
GPT-5.6 Luna wins on classification, data extraction, support chat, code review, summarisation, agent loop, translation, long-document qa, cached rag, nightly batch, while Claude Haiku 4.5 wins on none of the ten workloads above.
Beyond price, the two differ on input rate ($0.20 vs $1); output rate ($1.20 vs $5); cached-input rate ($0.02 vs $0.10); cache-write rate ($0.25 vs $1.25); context window (1,050,000 vs 200,000); long-context tier (yes, above 272,000 vs none).
Batch mode is available on GPT-5.6 Luna and Claude Haiku 4.5, which halves the nightly job. Both publish cached-input discounts, so a stable system prompt is worth caching on either.
FAQ
Is GPT-5.6 Luna cheaper than Claude Haiku 4.5?
On a blended 1M-input-plus-250K-output workload, GPT-5.6 Luna is cheaper by 78% ($0.50 vs $2.25). The gap changes for cache-heavy prompts and for batch jobs, as the workload table shows.
Which handles longer prompts, GPT-5.6 Luna or Claude Haiku 4.5?
Claude Haiku 4.5 has the larger context window (1,050,000 vs 200,000 tokens). Watch the long-context threshold on GPT-5.6 Luna: rates rise once input passes it.
Do both models support prompt caching?
GPT-5.6 Luna bills cached reads at $0.02 per 1M; Claude Haiku 4.5 at $0.10.
Related comparisons
- GPT-6 Luna vs GPT-5.6 Luna
- GPT-5.6 Luna vs GPT-5.4 nano
- Claude Sonnet 5 vs Claude Haiku 4.5
- All OpenAI pricing
- All Anthropic pricing