Claude Opus 5.5 API Pricing: Up to 67% Off, Model ID and Agent Guide
Claude Opus 5.5 API pricing is up to 67% below Anthropic's reference price on APIMaster. See claude-opus-5-5, 1M context, $4/$20 official pricing, live routes from $1.3359/$6.6794, and agent guidance.
Published 2026-09-21 · Updated 2026-09-23
Claude Opus 5.5 is Anthropic's released high-end model for complex agent work, coding, long tasks, and knowledge-work deliverables. Its API model ID is claude-opus-5-5. It has a 1M-token context window, up to 128K output tokens, a June 2026 knowledge cutoff, and always-on adaptive thinking with medium as the default behavior.
Anthropic lists $4 input / $20 output per 1M tokens, $0.20 cache reads, and $5 five-minute cache writes. That cache-read price matters for agent loops that repeatedly reuse a large context. Anthropic says typical workloads can cost about 40% less than Opus 5, but that is not a universal guarantee: long, high-effort jobs can still produce a great many output tokens.
APIMaster has five live, fingerprint-checked claude-opus-5-5 routes, from $1.3359 input / $6.6794 output per 1M tokens, up to ~67% off Anthropic's $4 / $20 reference price, checked September 23, 2026. Open the Claude Opus 5.5 marketplace card for the current routes and prices; the card is authoritative. One APIMaster key covers Claude, GPT, Gemini, DeepSeek, Kimi, and other listed providers.
| Model ID | Reference price (per 1M input / output) | Current APIMaster price | Active routes | Discount | Model card |
|---|---|---|---|---|---|
claude-opus-5-5 |
$4 / $20 |
$1.3359 / $6.6794 |
6 | 67% off | View discount |
First published September 21; updated September 23, 2026.
GPT Trial(GPT-6 Astra, GPT-6 Sol, GPT-5.6, and GPT-5.5 available)
Sign up and claim a $20 GPT trial
Claude Opus 5.5 specifications
| Item | Claude Opus 5.5 |
|---|---|
| API model ID | claude-opus-5-5 |
| Positioning | High-end Claude tier for demanding coding, agent, and knowledge-work tasks |
| Context window | 1M tokens |
| Maximum output | 128K tokens |
| Knowledge cutoff | June 2026 |
| Reasoning | Adaptive thinking, always enabled; default behavior is medium |
| Access | Claude app, Claude Code, API, AWS, GCP, and Azure |
Opus 5.5 is not a rumor, a preview label, or an alias to configure speculatively. It is a released model with a published ID and pricing. Anthropic also raised five-hour allowances for Pro, Max, and Team plans at launch; availability and limits are product-specific, so check the current provider documentation for your plan.
Claude Opus 5.5 API pricing
All figures are USD per 1M tokens.
| Model | Input | Output | Cache read | Cache write (5 min) |
|---|---|---|---|---|
| Claude Opus 5.5 | $4.00 | $20.00 | $0.20 | $5.00 |
| Claude Opus 5 | $5.00 | $25.00 | $0.50 | $6.25 |
Fast mode is $8 input / $40 output per 1M tokens and Anthropic says it can be up to about 2.5x faster. Unlike some long-context pricing schedules, Opus 5.5's 1M context does not add a separate long-context price tier.
APIMaster pricing and route availability
| Model | Live routes | APIMaster from | Official reference | Saving | Card |
|---|---|---|---|---|---|
| Claude Opus 5.5 | 5 | $1.3359 / $6.6794 | $4.00 / $20.00 | up to ~67% off | Opus 5.5 |
Prices and active route count were checked September 23, 2026. “From” is the lowest active route, not the only route. Check the market card before committing volume.
What is Claude Opus 5.5 good at?
Use Opus 5.5 for work where one strong attempt and sustained context are worth more than minimizing every token:
- long-horizon coding agents and complex repository changes;
- multi-step research, analysis, and knowledge-work deliverables;
- tasks that need a long-running tool loop, review, and self-correction;
- high-value workflows where rework, missed details, or an incomplete delivery costs more than the higher model bill.
Its low cache-read price is especially relevant when an agent keeps a large project context warm. Still measure the whole task: an agent that writes a very long answer can cost more than the input-rate headline suggests.
Capability evidence and limits
Anthropic reports 66.4% on Terminal-Bench 4.0, 54.4% on FrontierCode v1.1, 57.8% on CursorBench 4.0, and 1846 Elo on GDPval-AA v2.1. Artificial Analysis lists an Opus 5.5 max Intelligence Index around 58, with an estimated $5.98 per task and roughly 119K output tokens per task.
These are useful signals, not a promise that Opus wins every benchmark or every workload. On AutomationBench and Terminal-Bench-Science, published comparisons do not put it ahead of GPT-6 Astra. Anthropic also says some safety-related tasks can be routed to a weaker model, which can affect benchmark results. Evaluate the work you actually ship.
Claude Opus 5.5 vs GPT-6 Sol
The cleanest distinction is quality ceiling versus task cost, not a one-number winner.
| Same third-party harness, max settings | GPT-6 Sol | Claude Opus 5.5 |
|---|---|---|
| Artificial Analysis Intelligence Index | About 48 | About 58 |
| Estimated cost per task | About $1.06 | About $5.98 |
| Output tokens per task | About 31K | About 119K |
| Official input / output rate | $2 / $10 | $4 / $20 |
| Cache read | $0.20 | $0.20 |
For a difficult long task that must land cleanly, Opus 5.5 is the better starting point. For large-scale loops, cost-sensitive coding, and a capable middle tier, GPT-6 Sol is usually the more economical choice. Do not read this as an official equivalence between Sol max and any Opus setting: use the same repository, tools, effort policy, and acceptance criteria when you evaluate them.
Use Claude Opus 5.5 on APIMaster
- Create an APIMaster account.
- Create an API key in the console.
- Set your OpenAI-compatible base URL to
https://apimaster.ai/v1. - Set
modeltoclaude-opus-5-5. - Measure first-pass acceptance, retry rate, cache hit rate, output length, and total task cost on your own workload.
APIMaster verifies listed model routes with fingerprint checks. Review the live Opus 5.5 card before production use.
FAQ
Is Claude Opus 5.5 released?
Yes. The published API model ID is claude-opus-5-5.
How much does Claude Opus 5.5 cost?
The official standard price is $4 input / $20 output per 1M tokens, with $0.20 cache reads and $5 five-minute cache writes. Fast mode is $8 / $40.
Does Opus 5.5 support 1M context without a long-context surcharge?
Anthropic lists a 1M context window without a separate long-context price tier. Standard input, output, cache, and Fast mode pricing still apply.
Is Claude Opus 5.5 cheaper than Opus 5?
Its standard input and output list prices are 20% lower, and cache reads are $0.20 instead of $0.50. Anthropic says typical workloads can be about 40% cheaper, but actual cost depends on cache hits, thinking, output length, and retries.
Should I choose Opus 5.5 or GPT-6 Sol?
Choose Opus 5.5 for the strongest long-task, coding-agent, and knowledge-work result you can justify. Choose GPT-6 Sol when a substantially lower task cost matters and a strong middle tier is sufficient. Test both on the same acceptance criteria.
Sources and verification date
- Anthropic platform documentation, checked September 23, 2026: model ID, context, output limit, knowledge cutoff, reasoning behavior, availability, and pricing.
- Artificial Analysis, checked September 23, 2026: third-party index, task-cost, and output estimates; results are harness-specific.
- APIMaster public marketplace data, checked September 23, 2026: active route count, prices, and fingerprint status.
Start with Claude Opus 5.5
Create an APIMaster account, create a key in the console, point your client to https://apimaster.ai/v1, and set model to claude-opus-5-5. Check the live Opus 5.5 marketplace card for current route availability and price before you send production traffic.