vs
5:1
overall capability wins (0 ties across 6 scored). Kimi K2.5 leads overall.
- Reasoning4:0
- Specs1:1
Verdict
Kimi K2.5 edges reasoning & knowledge; Kimi Linear 48B A3B Instruct offers ~4.1x larger context (1.0M vs 256K).
- Kimi K2.5 edges reasoning & knowledge
- Kimi Linear 48B A3B Instruct offers ~4.1x larger context (1.0M vs 256K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Kimi K2.5
- $1.23
- Kimi Linear 48B A3B Instruct
- -
Price per 1M tokens
Full comparison
Organization
Moonshot
Moonshot
Family
Kimi
Kimi
License
Modified MIT
MIT
Open weights
Yes
Yes
Release
Jan 27, 2026
Oct 30, 2025
Knowledge cutoff
-
-
API / provider
Moonshot
No primary API price listed
Modalities
text, image → text
text → text
Specs
Context window
256K
1.0M
Max output
64K
-
Parameters
1.0T (32B act.)
48B (3B act.)
Pricing
Input $/1M
$0.60
—
Output $/1M
$2.50
—
Blended $/1M (3∶1)
$1.07
-
Speed
tok/s
70
-
TTFT (s)
0.45
-
Reasoning
MMLU-Pro
87.1
—
GPQA Diamond
87.9
41.2
Humanity's Last Exam
30.7
2.5
AIME 2025
96.1
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
35.2
—
AA-LCR v1.1
78.0
28.0
CritPt
3.1
—
MMMU-Pro
75.4
—
IFBench
70.2
28.1
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
76.8
—
SWE-bench Pro
50.7
—
SWE-bench Multilingual
67.3
—
LiveCodeBench
85.0
—
Terminal-Bench 2.1
45.7
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
49.0
—
CursorBench
—
—
SWE-Rebench
49.0
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Kimi K2.5 | Kimi Linear 48B A3B Instruct | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Moonshot | Moonshot | — |
| Family | Kimi | Kimi | — |
| License | Modified MIT | MIT | — |
| Open weights | Yes | Yes | — |
| Release | Jan 27, 2026 | Oct 30, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Moonshot | No primary API price listed | — |
| Modalities | text, image → text | text → text | — |
| Specs | |||
| Context window | 256K | 1.0M | B +793K |
| Max output | 64K | - | — |
| Parameters | 1.0T (32B act.) | 48B (3B act.) | A +952 |
| Pricing | |||
| Input $/1M | $0.60 | — | — |
| Output $/1M | $2.50 | — | — |
| Blended $/1M (3∶1) | $1.07 | - | — |
| Speed | |||
| tok/s | 70 | - | — |
| TTFT (s) | 0.45 | - | — |
| Reasoning | |||
| MMLU-Pro | 87.1 | — | — |
| GPQA Diamond | 87.9 | 41.2 | A +46.7 pts |
| Humanity's Last Exam | 30.7 | 2.5 | A +28.2 pts |
| AIME 2025 | 96.1 | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 35.2 | — | — |
| AA-LCR v1.1 | 78.0 | 28.0 | A +50.0 pts |
| CritPt | 3.1 | — | — |
| MMMU-Pro | 75.4 | — | — |
| IFBench | 70.2 | 28.1 | A +42.1 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | 76.8 | — | — |
| SWE-bench Pro | 50.7 | — | — |
| SWE-bench Multilingual | 67.3 | — | — |
| LiveCodeBench | 85.0 | — | — |
| Terminal-Bench 2.1 | 45.7 | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 49.0 | — | — |
| CursorBench | — | — | — |
| SWE-Rebench | 49.0 | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Kimi K2.5
- Context
- 256K / 64K out
- Parameters
- 1.0T (32B act.)
- Price
- $0.60 / $2.50
- Speed
- 70 tok/s · 0.45s TTFT
- Modalities
- text, image → text
- License
- Modified MIT
Prior Kimi open MoE generation - strong long-context agentic coding before K3.
Kimi Linear 48B A3B Instruct
- Context
- 1.0M
- Parameters
- 48B (3B act.)
- Price
- —
- Speed
- —
- Modalities
- text → text
- License
- MIT
Moonshot AI's hybrid linear-attention MoE (48B total / 3B active) pairing Kimi Delta Attention with sparse global attention for cheaper long-context (up to 1M token) inference; no hosted API yet.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.