vs
0:3
overall capability wins (0 ties across 3 scored). Cursor Grok 4.5 leads overall.
- Coding0:1
- Pricing3:0
- Speed1:1
- Specs0:2
Verdict
Cursor Grok 4.5 leads coding (widest gap: +10.6 on CursorBench); Cursor Grok 4.5 offers ~2.5x larger context (500K vs 200K); Cursor Grok 4.5 allows ~2.0x more max output (128K vs 64K).
- Cursor Grok 4.5 leads coding (widest gap: +10.6 on CursorBench)
- Cursor Grok 4.5 offers ~2.5x larger context (500K vs 200K)
- Cursor Grok 4.5 allows ~2.0x more max output (128K vs 64K)
- Composer 2.5 is ~3.0x cheaper on a blended token basis than Cursor Grok 4.5
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Composer 2.5
- $1.13
- Cursor Grok 4.5
- $3.50
Composer 2.5 is about 3.1x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Cursor
Cursor
Family
Composer
Grok
License
Proprietary
Proprietary
Open weights
No
No
Release
May 18, 2026
Jul 9, 2026
Knowledge cutoff
-
-
API / provider
Cursor
Cursor
Modalities
text → text
text, image → text
Specs
Context window
200K
500K
Max output
64K
128K
Parameters
—
—
Pricing
Input $/1M
$0.50
$2
Output $/1M
$2.50
$6
Blended $/1M (3∶1)
$1
$3
Speed
tok/s
90
95
TTFT (s)
0.35
0.4
Reasoning
MMLU-Pro
—
—
GPQA Diamond
—
—
Humanity's Last Exam
—
—
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
—
AA-LCR v1.1
—
—
CritPt
—
—
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
79.8
—
LiveCodeBench
—
—
Terminal-Bench 2.1
—
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
—
—
CursorBench
56.1
66.7
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Composer 2.5 | Cursor Grok 4.5 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Cursor | Cursor | — |
| Family | Composer | Grok | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | May 18, 2026 | Jul 9, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Cursor | Cursor | — |
| Modalities | text → text | text, image → text | — |
| Specs | |||
| Context window | 200K | 500K | B +300K |
| Max output | 64K | 128K | B +64K |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $0.50 | $2 | A +$1.50 |
| Output $/1M | $2.50 | $6 | A +$3.50 |
| Blended $/1M (3∶1) | $1 | $3 | A +$2 |
| Speed | |||
| tok/s | 90 | 95 | B +5.00 |
| TTFT (s) | 0.35 | 0.4 | A +0.05 |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | — | — | — |
| Humanity's Last Exam | — | — | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | — | — |
| AA-LCR v1.1 | — | — | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | 79.8 | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | — | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | — | — | — |
| CursorBench | 56.1 | 66.7 | B +10.6 pts |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
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.
Cursor Grok 4.5
- Context
- 500K / 128K out
- Parameters
- —
- Price
- $2 / $6
- Speed
- 95 tok/s · 0.4s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Cursor's flagship - jointly trained with SpaceXAI on Cursor agent data; effort levels (low/medium/high) for hard long-running work. Distinct from the public xAI Grok 4.5 API SKU.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.