Analysis
Anthropic launched Claude Haiku 5.5 on Wednesday, the third new model in its Claude 5.5 family in the past month, pricing the update at $0.10 per million input tokens and $0.50 per million output tokens for prompts under 100,000 tokens — a cut of roughly 75% from Haiku 4.5's list price. The company says Haiku 5.5 is built for high-volume, lower-complexity work: classification, summarization, extraction, live customer support, voice agents and in-app assistants, according to a Reuters report.
A third ship in a month, ahead of a planned IPO
Haiku 5.5 follows Opus 5.5 and Sonnet 5.5, both released in the prior several weeks, rounding out the 5.5 generation of Anthropic's usual three-tier lineup — a flagship (Opus), a mid-tier workhorse (Sonnet), and a cheap, fast option (Haiku). Reuters reports the release comes as Anthropic moves ahead with a planned IPO, though the company hasn't disclosed listing timing or terms publicly. Haiku has historically been Anthropic's highest-volume model by raw token count even though it generates the least revenue per call, since it's the tier most developers reach for once cost, not raw capability, decides the build — which is exactly why Pulse has tracked Anthropic's product cadence closely as it heads toward a listing.
“A 75% price cut on the cheapest model in the lineup is as much a defensive move to hold share in that commodity layer as it is a feature release.”
Reuters also describes Haiku 5.5 as Anthropic's first Haiku release with built-in safeguards for "a narrow set of high-risk cybersecurity requests," while noting the company says most everyday tasks go unaffected. That's a meaningful product decision: baking request-level restrictions into the cheapest, highest-volume tier is where abuse-at-scale would show up first if it were going to.
Competing on price in a commodity layer
The $0.10/$0.50 pricing puts Haiku 5.5 in the same bracket where OpenAI's smaller model tiers and Google's Gemini Flash line already compete for high-volume inference contracts — a segment with thin margins and low switching costs, since most "cheap tier" models get accessed through the same handful of API gateways and orchestration layers. A 75% price cut on the cheapest model in the lineup is as much a defensive move to hold share in that commodity layer as it is a feature release.
What the headline misses: a price cut announced at launch is not the same as a price cut that holds. Anthropic, like its rivals, has adjusted list prices after initial launches before as usage patterns and margin pressure became clearer post-launch — worth watching whether the $0.10/$0.50 rate survives contact with real enterprise volume. It's also unclear how the new cybersecurity safeguards affect latency or false-positive rates for legitimate security tooling built on top of Haiku, a detail that matters more to security vendors than to the average chatbot integrator.
The next price move
Anthropic said in its Opus 5.5 launch materials that Sonnet 5.5 and Haiku 5.5 would ship "in the coming weeks" with similar gains in performance, efficiency and safety — this release closes that loop. The near-term tell is whether OpenAI or Google answer with their own cheap-tier price cuts within the next few weeks, the pattern that has held through 2026's model-pricing cycles; if neither moves, it suggests Anthropic is pricing to grab volume share rather than reacting to a live price war.
