Claude Haiku 5.5 Matches GPT-6 Luna's Price on Short Prompts — and Sonnet 5.5's Cache Reads Get Halved
Anthropic's Claude Haiku 5.5 costs $0.10/$0.50 per million input/output tokens for prompts up to 100,000 tokens, 90% below Haiku 4.5 and level with GPT-6 Luna's API list price; above 100k it is $0.50/$2.50. Artificial Analysis scores Haiku 5.5 at 43 and Luna at 38 on its Intelligence Index, both at max effort. Anthropic's own benchmark lead over Luna is vendor-reported and swings with effort: Terminal-Bench is about 39% at max, about 20% at the default medium setting. Anthropic also cut Sonnet 5.5 cache reads to $0.10.
Anthropic has priced its new small model at exactly GPT-6 Luna's list rate, for the agent workloads where small models burn most of the tokens. Claude Haiku 5.5, released October 7, costs $0.10 per million input tokens and $0.50 per million output for prompts up to 100,000 tokens. That matches GPT-6 Luna's API list price, which VentureBeat says it matches on all four lower-tier rates. It landed within the same hour OpenAI announced that GPT-6 Luna would reach ChatGPT's free users from October 8.
On the one independent board checked, Haiku 5.5 also scores higher at matched effort: 43 against Luna's 38 on Artificial Analysis's Intelligence Index, both at max. That is a single composite index, not a general verdict. There are two catches. The price match ends at 100k-token prompts, and the benchmark lead depends heavily on which effort setting a buyer actually runs.
Pricing
| Per million tokens | Input | Output |
|---|---|---|
| Claude Haiku 4.5 | $1.00 | $5.00 |
| Haiku 5.5, prompts ≤100k tokens | $0.10 | $0.50 |
| Haiku 5.5, prompts >100k tokens | $0.50 | $2.50 |
| GPT-6 Luna (list, short prompts) | $0.10 | $0.50 |
The short-prompt tier is a 90% cut on Haiku 4.5; the long-prompt tier is a 50% cut and costs five times the short one. Anthropic says the average cost of running Haiku 5.5 is about 75% lower than Haiku 4.5 once a new tokenizer is accounted for, because it uses slightly more tokens per task. VentureBeat reports that Luna's own long-context surcharge starts above 272,000 input tokens; on that figure, prompts between 100,000 and 272,000 tokens cost five times as much on Haiku 5.5 as on Luna at list.
Haiku 5.5 is on the Claude Platform as claude-haiku-5-5 and on AWS, Google Cloud and Microsoft Azure. Anthropic pitches it for summaries, compaction, classification and database queries, and as a subagent under Opus 5.5 and Sonnet 5.5. Anthropic calls it its fastest model at standard speed; its own footnote says Opus in Fast Mode is faster.
Effort changes the scorecard
Haiku 5.5 is the first Haiku with an adjustable effort setting, and VentureBeat reports medium is the default. That matters for every number below.
| Anthropic-reported | Haiku 5.5 | GPT-6 Luna |
|---|---|---|
| OSWorld 2.1 (offline subset) | 72.4% | 48.9% |
| Terminal-Bench 4.0 | 39.2% | 16.4% |
Anthropic's table puts Haiku 5.5 ahead of Luna on every eval where both are scored and behind Sonnet 5.5. VentureBeat notes the roughly 39% Terminal-Bench result is at maximum effort; at medium, the default, Haiku 5.5 scores about 20%. The effort levels used for the Luna column are not clear from what we read, so the vendor table is not a like-for-like comparison.
Artificial Analysis, checked at 02:56 UTC on October 8, does compare like with like on one composite: Haiku 5.5 (max) 43, GPT-6 Luna (max) 38. At lower settings Haiku drops quickly: 41 at xhigh, 38 at high, 34 at medium and 29 at low. A team running Haiku at its default setting is not buying the headline numbers.
The Sonnet cut may matter more
Anthropic also cut Sonnet 5.5 cache reads from $0.20 to $0.10 per million tokens, which it says makes most agentic work on Sonnet about 20% cheaper. For customers already running long agent loops on Sonnet, that cut may matter more to their bills than Haiku itself. Separately, Anthropic says Max and Team subscribers will receive monthly Claude Platform API credits starting this week ($100 for Max 5x, $200 for Max 20x, up to $500 pooled for Team), and its Python and TypeScript SDKs add computer-use and browser-use support in beta.
On safety, Anthropic says its alignment evaluations show far fewer misaligned behaviours than Haiku 4.5. It says Haiku's cyber safeguards permit more defensive work than Sonnet 5.5's but still block penetration testing. That sits alongside the company's new cyber verification tiers. Customer endorsements from Asana, HubSpot, Box, AlphaSense, Rogo and Cognition were supplied by Anthropic.
Limits
- Benchmark tables are Anthropic's. The Artificial Analysis comparison is independent but is one index at one snapshot.
- The price match with Luna holds only below 100,000 prompt tokens. Luna's 272k surcharge threshold is VentureBeat's figure; OpenAI's pricing page was not checked.
- Haiku 5.5 was not yet listed on LMArena's text leaderboard at a 02:56 UTC check on October 8, so there is no crowd-preference ranking yet.
Sources
- Anthropic: Introducing Claude Haiku 5.5 (October 7, 2026)
- Claude on X: "Introducing Claude Haiku 5.5…" (October 7, 2026)
- VentureBeat: Anthropic launches Claude Haiku 5.5 with 90% API price reduction, matching GPT-6 Luna (October 7, 2026)
- SiliconANGLE: Anthropic releases Claude Haiku 5.5 small model and halves Sonnet 5.5 cache read prices (October 7, 2026)
- Artificial Analysis: LLM Leaderboard (snapshot October 8, 2026, 02:56 UTC)
- LMArena: Text leaderboard (checked October 8, 2026, 02:56 UTC)
- Times of AI: GPT-6 rolls into ChatGPT (October 8, 2026)
- Times of AI: Claude Opus 5.5 (September 22, 2026)
- Times of AI: Claude Sonnet 5.5 (September 28, 2026)
- Times of AI: Anthropic's cyber verification tiers (October 7, 2026)