Claude Code
Last updated: April 22, 2026
Specs
| Vendor | Anthropic |
|---|
| Released | February 24, 2025 |
|---|
| Context window | 1M tokens |
|---|
| Input | $15 / M tokens |
|---|
| Output | $75 / M tokens |
|---|
Benchmarks
| mmlu | 92.3 |
|---|
| gpqa | 65.2 |
|---|
| humanEval | 93.1 |
|---|
| arenaElo | 1420 |
|---|
| sweBench | 64.3 |
|---|
| sweBenchPro | 64.3 |
|---|
| liveCodeBench | 75 |
|---|
| aiderPolyglot | 79 |
|---|
| terminalBench | 55 |
|---|
| mmluPro | 86 |
|---|
| hle | 22 |
|---|
| aime | 88 |
|---|
| math500 | 96 |
|---|
Strengths
- Ships enterprise SaaS MVPs in weeks, not months, with Opus 4.7 under the hood
- Edits files, runs tests and opens PRs from the terminal without context switching
- Hooks, MCP servers and slash commands that map onto any B2B stack (Spring, Django, Next.js, React Native)
Weaknesses
- Opus 4.7 list price adds up on multi-week SaaS builds: budget API credits accordingly
- Terminal-first UX asks non-technical founders to pair with a senior engineer
- Private cloud runs require an Anthropic subscription
Best for
- Building custom SaaS and B2B MVPs for companies from zero
- Internal tools and admin dashboards for mid-sized enterprises
- Long-running agent sessions that refactor legacy codebases ahead of a modernization project
Compare with other AI models