Claude Sonnet 5 is Anthropic's Sonnet-class model released June 30, 2026, positioned as its most agentic Sonnet yet. It plans, uses tools like browsers and terminals, and carries multi-step tasks to completion — often finishing work where earlier Sonnet models would stop partway. On Anthropic's own benchmarks its agentic performance lands close to Opus 4.8 while costing less per token, which is why the Sonnet 5 API is aimed squarely at developers running agents and high-volume coding workloads on a budget.
On GPTProto you call the same claude-sonnet-5 model string through one API key and one balance, at a price below Anthropic's direct rate — and the same key already works across 200+ other models if you want to route cheaper tasks elsewhere.
Sonnet 5 Specifications
| Spec | Claude Sonnet 5 |
|---|---|
| Provider | Anthropic |
| Model string | claude-sonnet-5 |
| Modality (on GPTProto) | Text → Text (no image input) |
| Context window | 1M tokens (standard pricing, no long-context surcharge) |
| Thinking effort | Adjustable (dial reasoning per request) |
| GPTProto price /1M | $1.6 input / $8 output |
| Anthropic price /1M | $2 / $10 (intro, through Aug 31 2026) → $3 / $15 standard |
| Endpoint |
|
Sonnet 5 vs Sonnet 4.6 vs Opus 4.8
All three are on GPTProto, so this compares real, callable models — not a placeholder baseline. Prices are GPTProto's; Anthropic list shown for reference.
| Sonnet 5 | Sonnet 4.6 | Opus 4.8 (reference) | |
|---|---|---|---|
| GPTProto input /1M | $1.6 | $2.4 | $4 |
| GPTProto output /1M | $8 | $12 | ~$20 |
| Anthropic list /1M | $2/$10 → $3/$15 | $3/$15 | $5/$25 |
| Agentic coding | 63.2% | 58.1% | 69.2% |
| Context window | 1M | 1M | 1M |
On GPTProto, Sonnet 5 is both the newer model and the cheaper one — it undercuts Sonnet 4.6 ($1.6/$8 vs $2.4/$12) while scoring higher on agentic coding. Opus 4.8 stays ahead on the hardest reasoning tasks; reach for it when accuracy matters more than cost, and route the rest to Sonnet 5.
One honest caveat: Sonnet 5 uses a newer tokenizer that generates about 30% more tokens for the same text than Sonnet 4.6. The per-token price is lower, but your token count per request is higher — so benchmark cost on your own prompts before assuming a flat 20% saving. Agentic-coding figures are Anthropic's reported numbers
Switching From the Official Anthropic API
Moving an existing Anthropic integration to the Sonnet 5 API on GPTProto is a drop-in change — you keep the same request format and model string:
- Base URL → point requests to
https://gptproto.com/v1/messages. - Auth → use
Authorization: Bearer YOUR_GPTPROTO_API_KEY(keep theanthropic-version: 2023-06-01header). - Model string → unchanged:
claude-sonnet-5.
Your message payloads stay the same, so most SDKs migrate by swapping two lines. Runnable Python/cURL is in the Quick Start above. The payoff: one key and one balance across 200+ models, no separate Anthropic account, and a per-token price below Anthropic's direct rate.
