Anthropic releases Claude Haiku 5.5, claims ~75% lower running cost than Haiku 4.5
Anthropic released Claude Haiku 5.5 on October 7, 2026. The company claims it costs around 75% less to run than Haiku 4.5 and is its fastest model to date. Anthropic also halved Claude Sonnet 5.5's cache read pricing and added a monthly API credit for Max and Team subscribers.
Claude Haiku 5.5 at a glance
How Claude Haiku 5.5 ranks now
| Benchmark | Score | Rank | Source |
|---|---|---|---|
| AA Intelligence Index | 36.5 index | 20 of 56 | Artificial Analysis, Nov 2026 ↗ |
| Arena Elo | 1421 Elo | 89 of 146 | LMArena, Nov 2026 ↗ |
| Speed (tok/s) | 257 tok/s | 8 of 105 | Artificial Analysis, Nov 2026 ↗ |
Rank is against every model in our database with a sourced score on that benchmark, each counted once with its best score.
Anthropic released Claude Haiku 5.5 on October 7, 2026, and claims the model costs around 75% less to run than Claude Haiku 4.5, which it replaces. Anthropic describes it as "the cheapest, fastest, and most capable small model we've ever released."
What Anthropic announced
Haiku 5.5 moves the small-model tier from version 4.5 to 5.5. According to 9to5Mac, the previous Haiku upgrade came about a year ago. It is the third Claude 5.5 model in roughly a month, following Opus 5.5 and Sonnet 5.5.
Anthropic positions Haiku 5.5 for "high-volume, cost-sensitive tasks." The company lists these target workloads:
- Summaries, compactions, database queries and classification requests
- Subagent work alongside Opus 5.5 and Sonnet 5.5 on coding tasks
- Speed-sensitive uses such as live customer support and browser use
Anthropic says Haiku 5.5 is its fastest model to date. No latency or throughput figures were included in the coverage reviewed.
Pricing and platform changes
The launch includes two other changes, according to Anthropic:
- Sonnet 5.5 cache reads are halved in price. Anthropic says this makes Sonnet 5.5 run around 20% cheaper on most agentic work.
- A new monthly API credit for Claude Max and Team subscribers, intended for building agents and applications on the Claude Platform.
The ~75% cost reduction, the ~20% agentic savings and the "most capable small model" description are Anthropic's own claims. They have not been independently verified.
Not yet disclosed
The coverage we reviewed does not include:
- Per-million-token input and output pricing: not yet disclosed
- Context window size: not yet disclosed
- Benchmark scores: not yet disclosed
- Parameter count and training data cutoff: not yet disclosed
Anthropic's website compares Haiku 5.5 to OpenAI's GPT-5.5 Luna, another low-cost model. We have not confirmed the figures behind that comparison.
Per 9to5Mac, Anthropic's lineup now runs from Haiku (smallest, cheapest) through Sonnet and Opus to a higher tier called Fable/Mythos, which is more capable and more expensive.
What this means
A 75% cost reduction, if it holds in production, matters most for teams running high-volume pipelines. Classification, summarization and compaction are workloads where per-token cost often decides whether a feature ships.
The subagent framing is also telling. Anthropic is pitching Haiku 5.5 as the cheap executor in multi-model coding setups, with Opus 5.5 or Sonnet 5.5 planning and Haiku handling the repetitive steps. The halved Sonnet 5.5 cache read price points the same way. Agentic workloads re-read large contexts repeatedly, so cache pricing dominates their bills.
The naming of GPT-5.5 Luna as the comparison point shows low-cost models are now a direct competitive front. Until Anthropic publishes token prices and benchmark results, the real savings and the capability gap against Luna can't be assessed. Developers should test on their own workloads and wait for the full pricing table before migrating from Haiku 4.5.
Related Articles
Anthropic mathematician claims Riemann Hypothesis proof with Claude; Lean community certifies it correct
Anthropic mathematician Levent Alpöge said on November 20 that he and Claude found a proof of the Riemann Hypothesis, published as a formal proof in Lean. On November 21, the Lean and Mathlib communities issued a joint statement certifying the proof as correct. No human-readable paper exists yet.
Anthropic mathematician Levent Alpöge claims Claude-assisted proof of the Riemann hypothesis, posts it in Lean
Levent Alpöge, a mathematician employed by Anthropic, says he has proven the Riemann hypothesis, first proposed in 1859, working with Claude. He posted the proof coded in Lean, and experts have yet to confirm that it is correct.
Gemini 4 Neon scores 61 on Artificial Analysis Intelligence Index, 3rd overall, at $3/$15 per 1M tokens
Google DeepMind has released Gemini 4 Neon, its fourth model in seven weeks. Artificial Analysis measures it at 61 on the Intelligence Index (xhigh), 3rd overall, with token pricing of $3/$15 per 1M input/output tokens and a 1M-token context window.
RadixArk releases DaisyRose-27B, lifting Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%
RadixArk released DaisyRose-27B, its first model, built on Qwen3.8-27B and trained with reinforcement learning on 1,200 curated agent tasks. According to RadixArk, it raises Terminal Bench 4.1 pass@1 from 5.6% to 12.0% and scores higher than the base model on all seven reported benchmarks. Weights ship under Apache 2.0 in BF16, FP8 and NVFP4.
Comments
Loading...