Claude Haiku 5.5 vs Claude Sonnet 5.5
Reading a 26.2K-token pull request costs $0.0524 on Claude Sonnet 5.5, 20× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Claude Sonnet 5.5 a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Claude Sonnet 5.5 is −14 pts (95% interval −32 pts to +4 pts).
Side by side
| Figure | Claude Haiku 5.5 | Claude Sonnet 5.5 |
|---|---|---|
| Price tier | Economy | Mid-range |
| Bugs caughtshare of known bugs | 64% | 77% (better) |
| False alarmsper clean PR | 1.4 (better) | 2.0 |
| Cost per reviewmeasured | $0.0038 (better) | $0.0571 |
| Cost per bug caught | $0.0064 (better) | $0.083 |
| Exact linecatches on the expected lines | 82% | 82% (better) |
| Review timemean | 15s | 11s (better) |
| Typical review calllist price | $0.0038 (better) | $0.0766 |
| Input / 1M tokens | $0.10 (better) | $2.00 |
| Output / 1M tokens | $0.50 (better) | $10.00 |
| Context window | 1M | 1M |
Head to head
The known bugs both were scored on, by who caught each in at least one run.
By language
| Model | All | TypeScript6 bugs | Go4 bugs | Java4 bugs | Rust4 bugs | C#3 bugs | JavaScript3 bugs | Kotlin3 bugs | PHP3 bugs | Python3 bugs | Ruby3 bugs | Swift3 bugs | C++2 bugs | C1 bug |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Claude Haiku 5.5 | 64% | 83% | 75% | 40% | 75% | 67% | 33% | 67% | 100% | 39% | 67% | 67% | 50% | 0% |
| Claude Sonnet 5.5 | 77% | 83% | 50% | 45% | 100% | 56% | 100% | 67% | 67% | 67% | 100% | 100% | 100% | 100% |
Of the 42 known bugs both were scored on, both caught 23 and neither caught 4. Claude Haiku 5.5 caught 5 bugs that Claude Sonnet 5.5 missed: 2 in Go, 1 in Java, 1 in Kotlin and 1 in PHP. Claude Sonnet 5.5 caught 10 bugs that Claude Haiku 5.5 missed: 2 in JavaScript, 1 in C, 1 in C++, 1 in Go, 1 in Java, 1 in Kotlin, 1 in Ruby, 1 in Rust and 1 in Swift.
On pull requests with no known bug, Claude Haiku 5.5 raised 1.4 false alarms per review and Claude Sonnet 5.5 2. A measured review cost $0.0038 on Claude Haiku 5.5, 93% less than $0.0571 on Claude Sonnet 5.5.
| Pull request | Language · kind | Claude Haiku 5.5 | Claude Sonnet 5.5 |
|---|---|---|---|
| doctrine/ormFix collection filtering API for IN/NOT IN comparisons that require type conversions | PHP · Logic | 1/1 | 0/1 |
| go-gitea/giteaperf(gitdiff): optimize inline diff highlighting using cache | Go · Error handling | 1/1 | 0/1 |
| iBotPeaches/Apktoolrefactor: Improve renaming/injection of resources in stripped APKs | Java · Security | 1/1 | 0/1 |
| ktorio/ktorKTOR-9171 Redesign ByteReadChannel.readUTF8Line API | Kotlin · Logic | 1/1 | 0/1 |
| traefik/traefikAdd the option to define custom cipher suites for backend serversTransport | Go · API misuse | 1/1 | 0/1 |
| c-ares/c-aresEvent Thread could stall when not notified of new queries on existing connections that are in a bad state | C · Resource handling | 0/1 | 1/1 |
| ggml-org/llama.cpprpc : cache and reuse compute graphs | C++ · Security | 0/1 | 1/1 |
| go-chi/chifeat: middleware.ClientIP, a replacement for middleware.RealIP | Go · Security | 0/1 | 1/1 |
| kotest/kotestFix DataClassEq comparator | Kotlin · Logic | 0/1 | 1/1 |
| mockito/mockitoAdd core API to enable Kotlin singleton mocking | Java · API misuse | 0/1 | 1/1 |
| nushell/nushellreorganize nushell configuration | Rust · Error handling | 0/1 | 1/1 |
| pinojs/thread-streamrefactor: replace setTimeout with Atomics.waitAsync | JavaScript · Concurrency | 0/1 | 1/1 |
| pointfreeco/swift-navigationUIKit: Fix switch presented child double dismiss | Swift · Logic | 0/1 | 1/1 |
| rails/railsReimplement Action Cable redis adapter with redis-client | Ruby · Logic | 0/1 | 1/1 |
| sindresorhus/p-throttleAdd `weight` option | JavaScript · Logic | 0/1 | 1/1 |
Defects caught over all runs: a PR reviewed three times with one known bug shows 3/3 when it was caught every time.
Every bug each was scored on: Claude Haiku 5.5 and Claude Sonnet 5.5.