Claude Claude Code OAuth (Anthropic /v1/messages + oauth-2025-04-20)
The gold standard for coding agents.
Anthropic's Claude family is the reference point everyone else gets compared against. Sonnet 4.6 and Opus 4.7 routinely top SWE-Bench Verified, and instruction-following on long contexts is notably tighter than the alternatives. We benchmark via the Claude Code subscription that ships with Claude Pro / Max — flat monthly fee, no per-token surprises.
Strengths
- Best-in-class coding (Claude 4.7 leads SWE-Bench Verified at 87.6%)
- Long-context, careful instruction-following
- Flat subscription: no per-token surprises
When to use it
- Coding, long-document analysis, careful refactors
- When instruction-following accuracy matters
- If you already pay for Claude Pro/Max, reuse the quota from the CLI
Subscription plans
| Plan | Price | Quota | Available |
|---|---|---|---|
| Pro | $20/mo | Claude Code with shared Pro limits | yes |
| Max | $100/mo | 5× Pro quota | yes |
| Max+ | $200/mo | 20× Pro quota + Guest Passes | yes |
Notes: Annual billing available with discount. Max tiers unlock /passes Guest Pass slots.
Referral: Anthropic does not run a public commission affiliate program. Max users get 3 single-use Guest Passes (7-day Pro trial) via the /passes command, and Anthropic occasionally runs invite-contests with sweepstakes prizes.
Models tested on Claude
Speed numbers below are specific to Claude's routing and hardware. The same model may appear on other providers' pages with different throughput.
| Model | Best tok/s | Avg tok/s | Runs | Success | Longest output (chars) |
|---|---|---|---|---|---|
| claude-haiku-4-5-20251001 | 119.8 | 94.5 | 4 | 100% | 1,527 |
| claude-opus-4-7 | 58.3 | 54.5 | 3 | 100% | 1,993 |
| claude-sonnet-4-6 | 56.7 | 54.4 | 3 | 100% | 2,089 |
| claude-sonnet-4-5 | 51.1 | 49.8 | 3 | 100% | 2,291 |
| claude-opus-4-5 | 50.9 | 50.2 | 3 | 100% | 1,502 |