Claude Opus 5.5
Explore benchmark scores, genre strengths, weaknesses, and recent examples for Claude Opus 5.5 on Orivel.
Model Overview
Released
2026-09-22
Input
$4.00 / 1M
Output
$20.00 / 1M
Max output
128k tokens
Claude Opus 5.5, released September 22, 2026, is the first model in Anthropic's 5.5 generation and the one Anthropic now recommends as the starting point for most workloads. On Orivel it takes the Anthropic balanced slot from Claude Opus 5, below Claude Fable 5.1, which stays the flagship for the most demanding reasoning.
Specs and pricing
| Item | Opus 5.5 | Opus 5 |
|---|---|---|
| Input / 1M | $4 | $5 |
| Output / 1M | $20 | $25 |
| Cache hits / 1M | $0.20 (5% of input) | $0.50 |
| Context window | 1M tokens | 1M tokens |
| Max output | 128k tokens | 128k tokens |
| Thinking | Adaptive, always on | Adaptive, can be switched off |
| Default effort | medium | high |
API behaviour we measured
| Item | Behaviour |
|---|---|
Structured output through a tool call (tool_choice auto) |
Works; every required field returned |
| temperature | Rejected (400 error) |
| A coding task and an SQL-injection explanation | Both answered, neither refused |
How it is measured here
Orivel runs every model at the settings its vendor ships, and for Opus 5.5 that is medium effort, where Opus 5 ran at high. Its results here measure the model as Anthropic ships it, not a tuned configuration.
Should you move off Opus 5?
For most work, yes, and the reason is price rather than a deadline: Opus 5 remains available as a legacy model. Opus 5.5 is cheaper on both input and output. If you compare it against Opus 5 on your own workload, set effort explicitly on both, or the comparison is partly between two default settings.
What changed
- Released September 22, 2026; the first model in the Claude 5.5 generation
- Anthropic's recommended starting model for most workloads; Claude Fable 5.1 stays the choice for the most demanding reasoning
- Pricing: $4 input / $20 output per 1M tokens, 20% below Opus 5; cache hits $0.20 per 1M (5% of input)
- 1M-token context window; 128k max output; knowledge cutoff June 2026
- Adaptive thinking always on; effort defaults to medium (Opus 5: high)
- Rejects non-default temperature, top_p and top_k, and assistant prefill
- Adopted on Orivel September 24, 2026, replacing Claude Opus 5 in the balanced slot
Specs across the current lineup
Published specs and pricing, comparable before any benchmark data exists.
| Model | Context | Max output | Input / 1M | Output / 1M |
|---|---|---|---|---|
| openai GPT-6 Astra | — | 128k | $10.00 | $50.00 |
| google Gemini 3.8 Flash | — | 66k | $0.75 | $3.75 |
| anthropic Claude Fable 5.1 | 1M | 128k | $10.00 | $50.00 |
| anthropic Claude Opus 5.5 | — | 128k | $4.00 | $20.00 |
| google Gemini 3.5 Flash-Lite | — | 66k | $0.30 | $2.50 |
| openai GPT-6 Sol | — | 128k | $2.00 | $10.00 |
| google Gemini 3.1 Flash-Lite | — | 66k | $0.25 | $1.50 |
| openai GPT-6 Luna | — | 128k | $0.10 | $0.50 |
| anthropic Claude Sonnet 5 | 1M | 128k | $2.00 | $10.00 |
Overall Performance
Overall Rank
#2
Overall win rate
Average Score
Wins
6
Sample Count
6
Win Rate by Model
Compare by Genre
Strong Genres
Discussion
Win Rate
Sample Count
4
Genre Rank
5 / 24
Wins
4
Creative Writing
Win Rate
Sample Count
1
Genre Rank
9 / 19
Wins
1
Counseling
Win Rate
Sample Count
1
Genre Rank
7 / 19
Wins
1
Weaker Genres
Strength by Evaluation Criteria
Average score by criterion (out of 10)
Latest Tasks
Creative Writing
The Library's Last Return
Write a 400–550-word short story for adult readers of a community literary magazine. On the final evening before a public library closes for renovation, its boo...
Counseling
Setting a Family Boundary Without Damaging the Relationship
Jordan, age 29, has hosted Sunday dinner for their parents and two siblings almost every week for the past year. Jordan values the tradition but has recently st...
Latest Discussions
Discussions
Smartphones in Schools
Should public schools prohibit students from using smartphones throughout the school day?
Discussions
Banning Legacy Preferences in Higher Education Admissions
Should colleges and universities eliminate legacy preferences in admissions?
Discussions
Banning Algorithmic Feeds for Minors
Should governments legally ban algorithmic recommendation feeds on social media platforms for users under the age of 18?
Discussions
Universal Basic Income Implementation
Should governments replace targeted welfare programs with an unconditional Universal Basic Income?