vs
0:2
overall capability wins (0 ties across 2 scored). Claude Opus 5.5 leads overall.
- Pricing0:3
- Specs0:2
Verdict
Claude Opus 5.5 offers ~5.0x larger context (1M vs 200K); Claude Opus 5.5 allows ~4.0x more max output (128K vs 32K); Claude Opus 5.5 is ~3.8x cheaper on a blended token basis than Claude Opus 4.
- Claude Opus 5.5 offers ~5.0x larger context (1M vs 200K)
- Claude Opus 5.5 allows ~4.0x more max output (128K vs 32K)
- Claude Opus 5.5 is ~3.8x cheaper on a blended token basis than Claude Opus 4
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Claude Opus 4
- $33.75
- Claude Opus 5.5
- $9
Claude Opus 5.5 is about 3.8x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Anthropic
Anthropic
Family
Claude Opus
Claude Opus
License
Proprietary
Proprietary
Open weights
No
No
Release
May 22, 2025
Sep 22, 2026
Knowledge cutoff
-
2026-06
API / provider
Anthropic
Anthropic
Modalities
text, image → text
text, image → text
Specs
Context window
200K
1M
Max output
32K
128K
Parameters
—
—
Pricing
Input $/1M
$15
$4
Output $/1M
$75
$20
Blended $/1M (3∶1)
$30
$8
Speed
tok/s
50
-
TTFT (s)
0.75
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
—
—
Humanity's Last Exam
—
61.4
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
67.7
AA-Omniscience Accuracy
—
—
AA-LCR v1.1
—
—
CritPt
—
—
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
64.4
Chartography (With Tools)
—
89.0
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
62.4
—
Terminal-Bench 2.1
—
—
Aider Polyglot
70.7
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
—
66.9
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
66.4
CursorBench 4.0
—
57.8
FrontierCode v1.1 (Main)
—
54.4
Arena
LMArena Elo
1,414
—
| Metric | Claude Opus 4 | Claude Opus 5.5 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Anthropic | Anthropic | — |
| Family | Claude Opus | Claude Opus | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | May 22, 2025 | Sep 22, 2026 | — |
| Knowledge cutoff | - | 2026-06 | — |
| API / provider | Anthropic | Anthropic | — |
| Modalities | text, image → text | text, image → text | — |
| Specs | |||
| Context window | 200K | 1M | B +800K |
| Max output | 32K | 128K | B +96K |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $15 | $4 | B +$11 |
| Output $/1M | $75 | $20 | B +$55 |
| Blended $/1M (3∶1) | $30 | $8 | B +$22 |
| Speed | |||
| tok/s | 50 | - | — |
| TTFT (s) | 0.75 | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | — | — | — |
| Humanity's Last Exam | — | 61.4 | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | 67.7 | — |
| AA-Omniscience Accuracy | — | — | — |
| AA-LCR v1.1 | — | — | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | 64.4 | — |
| Chartography (With Tools) | — | 89.0 | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 62.4 | — | — |
| Terminal-Bench 2.1 | — | — | — |
| Aider Polyglot | 70.7 | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | — | 66.9 | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | 66.4 | — |
| CursorBench 4.0 | — | 57.8 | — |
| FrontierCode v1.1 (Main) | — | 54.4 | — |
| Arena | |||
| LMArena Elo | 1,414 | — | — |
Claude Opus 4
- Context
- 200K / 32K out
- Parameters
- —
- Price
- $15 / $75
- Speed
- 50 tok/s · 0.75s TTFT
- Modalities
- text, image → text
- License
- Proprietary
First Claude 4 Opus generation — hybrid reasoning and strong agentic coding before 4.1 / 4.5 / 5.
Claude Opus 5.5
- Context
- 1M / 128K out
- Parameters
- —
- Price
- $4 / $20
- Speed
- —
- Modalities
- text, image → text
- License
- Proprietary
Anthropic's September 2026 Opus model for long-running agentic coding and knowledge work. It has a 1M context window, 128K output limit, always-on adaptive thinking, and standard API rates of $4/$20 per 1M input/output tokens.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.
These two models have no overlapping published benchmarks in our dataset. Compare specs and pricing instead.