No-BS Take · Pricing

GPT-6 Luna vs Claude Haiku 5.5: Same Price Until Your Prompt Passes 100K Tokens

The numbers

GPT-6 Luna, OpenAI (per 1M tokens)Official price$0.10 / $0.50 up to 272K input tokens; $0.20 / $0.75 above; cached input $0.01 and $0.02OpenAI API pricingChecked Nov 8, 2026 by AISelect editorial research, high confidence
Claude Haiku 5.5, Anthropic (per 1M tokens)Official price$0.10 / $0.50 for prompts up to 100K tokens; $0.50 / $2.50 above; cache reads $0.01 and $0.05Claude pricingChecked Nov 8, 2026 by AISelect editorial research, high confidence
Gemini 3.8 Flash, Gemini API (per 1M tokens)Official price$0.75 / $3.75 through Dec 31, 2026; $1.50 / $7.50 from Jan 1, 2027Gemini API pricingChecked Nov 8, 2026 by AISelect editorial research, high confidence
GPQA Diamond (Epoch AI, reasoning setting in brackets)Third-party benchmarkGemini 3.8 Flash 95.4% (high); GPT-6 Luna 90.5% (max); Claude Haiku 5.5 89.6% (max)Epoch AI benchmarks (CC BY 4.0)Checked Nov 8, 2026 by AISelect daily benchmark tracking (Epoch AI), medium confidence
Monthly cost for 10M input + 2M output tokensOur calculationPrompts up to 100K: Luna $2.00, Haiku 5.5 $2.00; 100K to 272K: Luna $2.00, Haiku 5.5 $10.00; above 272K: Luna $3.50, Haiku 5.5 $10.00; Gemini 3.8 Flash $15.00Our calculation from the listed pricesChecked Nov 8, 2026 by AISelect editorial research, high confidence

Frozen at publication. Current prices are shown below.

GPT-6 Luna and Claude Haiku 5.5 cost exactly the same: $0.10 per million input tokens, $0.50 per million output tokens, and $0.01 for cache reads. That holds until a prompt passes 100,000 tokens. From there Anthropic charges $0.50 / $2.50 for Haiku 5.5, five times as much. OpenAI keeps Luna at $0.10 / $0.50 up to 272K input tokens and charges $0.20 / $0.75 above that.

If your prompts are short, pick either on quality. If they are long, the choice is made for you.

The numbers

One month of 10 million input tokens and 2 million output tokens, by the size of each prompt:

Prompt sizeGPT-6 LunaClaude Haiku 5.5Gemini 3.8 Flash, until Dec 31
Up to 100K tokens$2.00$2.00$15.00
100K to 272K tokens$2.00$10.00$15.00
Above 272K tokens$3.50$10.00$15.00
GPQA Diamond90.5% (max)89.6% (max)95.4% (high)

Prices from the official OpenAI, Anthropic and Google pricing pages, checked on November 8, 2026. GPQA Diamond scores from Epoch AI at the reasoning setting shown. The cost columns are our arithmetic and ignore caching and batch discounts.

All three accept prompts far longer than these thresholds: our data lists 922K tokens of context for Luna and about 1M for Haiku 5.5 and Gemini 3.8 Flash. Our model pages show each model's standard rate; the long-prompt rates live on the providers' pages linked below.

Who crosses 100K without noticing

The threshold applies to each request, not to your monthly total. A few common setups cross it quietly:

  • Agents and coding tools that resend the conversation, file contents and tool output with every step. A long session passes 100K tokens per call well before anyone looks at the bill.
  • Document workflows that put a whole contract, report or codebase into one prompt instead of retrieving the relevant parts.
  • Long chat histories that are never trimmed or summarised.

Caching does not escape the tier: Haiku 5.5's cache reads cost $0.05 per million above 100K tokens instead of $0.01. Luna's cached input goes from $0.01 to $0.02 above 272K.

Quality is a tie, so price decides

On GPQA Diamond, Luna (90.5%) and Haiku 5.5 (89.6%) are within one point, about two of 198 questions. Neither is the model to pick for the hardest reasoning. Gemini 3.8 Flash scores 95.4% for seven and a half times the short-prompt price, and Google doubles its rate to $1.50 / $7.50 on January 1, 2027, which makes the same month $30.

What I would do

  • Short prompts (chat, classification, extraction): the price is identical. Run both on your own examples and keep the better one.
  • Prompts between 100K and 272K tokens: GPT-6 Luna costs a fifth of Haiku 5.5. If you want to stay on Haiku, keep requests under 100K with retrieval or summarised history.
  • Prompts above 272K: Luna still costs about a third of Haiku 5.5 ($3.50 against $10.00 for the month above).
  • Answers are not good enough at this tier: test Gemini 3.8 Flash, and budget for its January price.

Compare the live prices in Claude Haiku 5.5 vs GPT-6 Luna and Gemini 3.8 Flash vs GPT-6 Luna, or see how the whole OpenAI and Anthropic price lists line up.

The prices behind this take

Checked by hand on November 8, 2026: OpenAI API pricing (GPT-6 Luna: $0.10 / $0.50 up to 272K input tokens, $0.20 / $0.75 above, cached input $0.01 and $0.02), Claude pricing (Haiku 5.5: $0.10 / $0.50 for prompts up to 100K tokens, $0.50 / $2.50 above, cache reads $0.01 and $0.05) and Gemini API pricing (Gemini 3.8 Flash: $0.75 / $3.75 until December 31, 2026, $1.50 / $7.50 from January 1, 2027). Accuracy scores: Epoch AI benchmarks, GPQA Diamond. Context windows: our daily tracked data.

The honest summary: same model tier, same score, same price, until your prompts get long. Check how big your requests really are before you choose.

Today's API prices

Updated daily from our price tracking; the numbers above are frozen at publication.

ModelCheapest providerInput / 1MOutput / 1MProviders
GPT-6 LunaOpenAI$0.10$0.504
Claude Haiku 5.5Anthropic$0.10$0.505
Gemini 3.8 FlashDatabricks$0.75$3.754

Opinion, with receipts

More No-BS Takes

This is an opinion piece. The numbers are sourced; the conclusions are the author's. No provider paid for or reviewed it.