This page is a free Claude Haiku 5.5 vs GPT-6 Luna comparison tool: an in-browser API cost calculator (tokens in/out, cached share, batch, prompt-length tiers) plus an official spec and vendor-reported benchmark table. Haiku 5.5 (claude-haiku-5-5, launched 7 Oct 2026) list price is $0.10 / $0.50 per MTok for prompts ≤100K tokens and $0.50 / $2.50 when the prompt is over 100K (whole request). GPT-6 Luna (gpt-6-luna) is $0.10 / $0.50, with 2× input/cache and 1.5× output when input exceeds 272K tokens. The calculator does not call either API.
What this Haiku 5.5 vs GPT-6 Luna page is
This free Toolsiftly page is a working API cost calculator and an official-spec comparison for Claude Haiku 5.5 (claude-haiku-5-5) and GPT-6 Luna (gpt-6-luna). It does not call either API. Numbers come from Anthropic and OpenAI list prices checked on 15 October 2026.
How the cost is calculated
- Haiku 5.5: prompts up to 100,000 tokens use $0.10 / $0.50 per MTok (input / output); prompts over 100,000 tokens pay $0.50 / $2.50 for the whole request. Cache read $0.01 / $0.05; 5-minute cache write $0.125 / $0.625; 1-hour write $0.20 / $1.00. Batch API −50%. Context 1M; max output 128K.
- GPT-6 Luna: $0.10 / $0.50; cached input $0.01; cache write $0.125. If input exceeds 272K tokens, OpenAI applies 2× to input and cache rates and 1.5× to output for the full request. Batch/Flex −50%; Fast 2×. Context 1.05M (922K max input); max output 128K; knowledge cutoff 18 May 2026.
Toggle “Charge cache write” when your first request stores a prompt prefix; leave it off for steady-state cache hits only. Sources: Claude Platform pricing, GPT-6 Luna model page.
How to use the calculator
- Enter uncached input tokens and output tokens, or tap a preset.
- Set cached share (%) if part of the prompt is a cache hit.
- Optionally raise the Haiku tokenizer ratio (third-party DataCamp figure ~1.7× is user-adjustable, not a claim we tested).
- Toggle Batch and/or Luna Fast, then read which model is cheaper for that mix.
Specs at a glance
| Claude Haiku 5.5 | GPT-6 Luna | |
|---|---|---|
| Model id | claude-haiku-5-5 |
gpt-6-luna |
| Maker | Anthropic | OpenAI |
| List input / output (short prompts) | $0.10 / $0.50 per MTok | $0.10 / $0.50 per MTok |
| Long-prompt pricing | Prompt >100K → $0.50 / $2.50 for the whole request | Input >272K → 2× input & cache, 1.5× output for the full request |
| Cache read | $0.01 (≤100K) / $0.05 (>100K) | $0.01 |
| Cache write | 5m $0.125 / $0.625; 1h $0.20 / $1.00 | $0.125 |
| Batch | −50% | Batch / Flex −50%; Fast 2× |
| Context | 1M | 1.05M (922K max input) |
| Max output | 128K | 128K |
| Knowledge cutoff | — (see Anthropic docs) | 18 May 2026 |
Prices checked 15 Oct 2026 against Claude Platform pricing and the GPT-6 Luna model page.
Vendor-reported benchmarks (Anthropic table)
Anthropic published these figures on the Haiku 5.5 launch page. They are vendor-reported — not independent third-party results. We label them as such and do not claim we re-ran the evals.
| Benchmark | Haiku 5.5 | GPT-6 Luna |
|---|---|---|
| GDPval-AA v2.1 | 1620 | 1437 |
| OSWorld 2.1 (offline subset) | 72.4% | 48.9% |
| Terminal-Bench 4.0 | 39.2% | 16.4% |
| FrontierCode 1.1 (Main) | 46.4% | 42.4% |
| Chartography (no tools) | 46.4% | 29.1% |
Source: Introducing Claude Haiku 5.5 (7 Oct 2026). Sonnet 5.5 and Haiku 4.5 columns appear on Anthropic’s page for context; this tool focuses on Haiku 5.5 vs Luna.
When each model tends to fit
- Haiku 5.5: high-volume, cost-sensitive work — summaries, classification, subagents, browser/computer-use helpers — especially when prompts stay under 100K tokens.
- GPT-6 Luna: OpenAI’s efficient high-volume model with a 1.05M context window and May 2026 knowledge cutoff; use Fast when latency matters more than price, Batch/Flex when it does not.
Tokenizers differ. A third-party write-up (DataCamp) reported roughly 1.7× more Haiku input tokens for the same text. That is not our measurement — the calculator’s “Haiku tokenizer ratio” lets you apply any factor yourself.
Frequently asked questions
What is Claude Haiku 5.5?
Anthropic’s fastest, most cost-efficient Claude model as of its 7 October 2026 launch. The API id is claude-haiku-5-5. It targets high-volume tasks and subagent work, with $0.10/$0.50 per MTok for prompts up to 100K tokens.
What is GPT-6 Luna?
OpenAI’s efficient GPT-6 model for focused, high-volume tasks. The API id is gpt-6-luna. List price is $0.10/$0.50 per MTok, with a 1.05M context window (922K max input), 128K max output, and a knowledge cutoff of 18 May 2026.
Which is cheaper, Haiku 5.5 or GPT-6 Luna?
For equal token counts on short prompts they share the same $0.10/$0.50 list rates. Differences come from prompt-length tiers (Haiku jumps at 100K; Luna at 272K), cache write vs read mix, Batch/Fast multipliers, and tokenizer token counts. Use the calculator above for your mix.
Do Haiku 5.5 benchmarks beat GPT-6 Luna?
On Anthropic’s vendor-reported table, Haiku 5.5 scores higher on GDPval-AA (1620 vs 1437), OSWorld 2.1 (72.4% vs 48.9%), Terminal-Bench 4.0 (39.2% vs 16.4%), FrontierCode (46.4% vs 42.4%) and Chartography (46.4% vs 29.1%). Those figures are Anthropic’s, not an independent lab run.
Does this tool call Anthropic or OpenAI?
No. It only multiplies the tokens you enter by the published list prices in your browser.
What about Haiku 5.5 benchmarks plural?
Same page — the benchmark section covers the official Anthropic comparison table for Haiku 5.5 (including vs GPT-6 Luna).