vs
3:0
overall capability wins (0 ties across 3 scored). Claude Opus 5 leads overall.
- Coding1:0
- Pricing0:3
- Speed0:2
- Specs2:0
Verdict
Claude Opus 5 leads coding (widest gap: +13.9 on CursorBench); Claude Opus 5 offers ~5.0x larger context (1M vs 200K); Claude Opus 5 allows ~2.0x more max output (128K vs 64K).
- Claude Opus 5 leads coding (widest gap: +13.9 on CursorBench)
- Claude Opus 5 offers ~5.0x larger context (1M vs 200K)
- Claude Opus 5 allows ~2.0x more max output (128K vs 64K)
- Composer 2.5 is ~10.0x cheaper on a blended token basis than Claude Opus 5
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Claude Opus 5
- $11.25
- Composer 2.5
- $1.13
Composer 2.5 is about 10.0x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Anthropic
Cursor
Family
Claude Opus
Composer
License
Proprietary
Proprietary
Open weights
No
No
Release
Jul 24, 2026
May 18, 2026
Knowledge cutoff
-
-
API / provider
Anthropic
Cursor
Modalities
text, image → text
text → text
Specs
Context window
1M
200K
Max output
128K
64K
Parameters
—
—
Pricing
Input $/1M
$5
$0.50
Output $/1M
$25
$2.50
Blended $/1M (3∶1)
$10
$1
Speed
tok/s
72
90
TTFT (s)
0.55
0.35
Reasoning
MMLU-Pro
—
—
GPQA Diamond
93.2
—
Humanity's Last Exam
54.9
—
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
60.9
—
AA-LCR v1.1
79.3
—
CritPt
29.1
—
MMMU-Pro
84.7
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
79.8
LiveCodeBench
—
—
Terminal-Bench 2.1
89.1
—
Aider Polyglot
—
—
Terminal-Bench 3
42.7
—
BigCodeBench
—
—
SciCode
56.4
—
CursorBench
70.0
56.1
SWE-Rebench
63.4
—
NL2Repo-Bench
—
—
DeepSWE
73.7
—
WebDev Arena
1,687
—
Terminal-Bench 4.0
51.8
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,493
—
| Metric | Claude Opus 5 | Composer 2.5 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Anthropic | Cursor | — |
| Family | Claude Opus | Composer | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | Jul 24, 2026 | May 18, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Anthropic | Cursor | — |
| Modalities | text, image → text | text → text | — |
| Specs | |||
| Context window | 1M | 200K | A +800K |
| Max output | 128K | 64K | A +64K |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $5 | $0.50 | B +$4.50 |
| Output $/1M | $25 | $2.50 | B +$22.50 |
| Blended $/1M (3∶1) | $10 | $1 | B +$9 |
| Speed | |||
| tok/s | 72 | 90 | B +18 |
| TTFT (s) | 0.55 | 0.35 | B +0.20 |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 93.2 | — | — |
| Humanity's Last Exam | 54.9 | — | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 60.9 | — | — |
| AA-LCR v1.1 | 79.3 | — | — |
| CritPt | 29.1 | — | — |
| MMMU-Pro | 84.7 | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | 79.8 | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 89.1 | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | 42.7 | — | — |
| BigCodeBench | — | — | — |
| SciCode | 56.4 | — | — |
| CursorBench | 70.0 | 56.1 | A +13.9 pts |
| SWE-Rebench | 63.4 | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | 73.7 | — | — |
| WebDev Arena | 1,687 | — | — |
| Terminal-Bench 4.0 | 51.8 | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,493 | — | — |
Claude Opus 5
- Context
- 1M / 128K out
- Parameters
- —
- Price
- $5 / $25
- Speed
- 72 tok/s · 0.55s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Anthropic's July 2026 flagship - top Artificial Analysis Intelligence Index and strong agentic coding at Opus 4.8 pricing.
Composer 2.5
- Context
- 200K / 64K out
- Parameters
- —
- Price
- $0.50 / $2.50
- Speed
- 90 tok/s · 0.35s TTFT
- Modalities
- text → text
- License
- Proprietary
Cursor's own agentic coding model (standard tier) — Cursor reports 69.3% Terminal-Bench 2.0 and 79.8% SWE-bench Multilingual; draws from the Cursor Models usage pool at $0.50/$2.50.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.