vs
6:0
overall capability wins (0 ties across 6 scored). K2 Horizon 375B A23B leads overall.
- Reasoning3:0
- Coding1:0
- Specs2:0
Verdict
K2 Horizon 375B A23B leads coding (widest gap: +14.3 on SciCode); K2 Horizon 375B A23B offers ~8.0x larger context (524K vs 66K).
- K2 Horizon 375B A23B leads coding (widest gap: +14.3 on SciCode)
- K2 Horizon 375B A23B offers ~8.0x larger context (524K vs 66K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- K2 Horizon 375B A23B
- -
- OLMo 3 32B Think
- $0.17
Price per 1M tokens
Full comparison
Organization
MBZUAI Institute of Foundation Models
Ai2
Family
Other
Other
License
Apache 2.0
Apache 2.0
Open weights
Yes
Yes
Release
Sep 3, 2026
Nov 1, 2025
Knowledge cutoff
-
-
API / provider
No primary API price listed
self-host
Modalities
text → text
text → text
Specs
Context window
524K
66K
Max output
-
8K
Parameters
375B (23B act.)
32B
Pricing
Input $/1M
—
$0.10
Output $/1M
—
$0.30
Blended $/1M (3∶1)
-
$0.15
Speed
tok/s
-
55
TTFT (s)
-
0.4
Reasoning
MMLU-Pro
—
75.9
GPQA Diamond
87.3
61.0
Humanity's Last Exam
32.0
6.4
AIME 2025
—
73.7
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
18.2
14.9
AA-LCR v1.1
80.0
—
CritPt
4.6
—
MMMU-Pro
—
—
IFBench
—
49.1
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
67.2
Terminal-Bench 2.1
71.9
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
42.9
28.6
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
1,307
| Metric | K2 Horizon 375B A23B | OLMo 3 32B Think | Delta |
|---|---|---|---|
| Identity | |||
| Organization | MBZUAI Institute of Foundation Models | Ai2 | — |
| Family | Other | Other | — |
| License | Apache 2.0 | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Sep 3, 2026 | Nov 1, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | No primary API price listed | self-host | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 524K | 66K | A +459K |
| Max output | - | 8K | — |
| Parameters | 375B (23B act.) | 32B | A +343 |
| Pricing | |||
| Input $/1M | — | $0.10 | — |
| Output $/1M | — | $0.30 | — |
| Blended $/1M (3∶1) | - | $0.15 | — |
| Speed | |||
| tok/s | - | 55 | — |
| TTFT (s) | - | 0.4 | — |
| Reasoning | |||
| MMLU-Pro | — | 75.9 | — |
| GPQA Diamond | 87.3 | 61.0 | A +26.3 pts |
| Humanity's Last Exam | 32.0 | 6.4 | A +25.6 pts |
| AIME 2025 | — | 73.7 | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 18.2 | 14.9 | A +3.3 pts |
| AA-LCR v1.1 | 80.0 | — | — |
| CritPt | 4.6 | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | 49.1 | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | 67.2 | — |
| Terminal-Bench 2.1 | 71.9 | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 42.9 | 28.6 | A +14.3 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | 1,307 | — |
K2 Horizon 375B A23B
- Context
- 524K
- Parameters
- 375B (23B act.)
- Price
- —
- Speed
- —
- Modalities
- text → text
- License
- Apache 2.0
Flagship model (375B MoE / 23B active) in the Institute of Foundation Models' K2 Horizon fleet, released fully open — weights, code, training data, and methodology — under Apache 2.0.
OLMo 3 32B Think
- Context
- 66K / 8K out
- Parameters
- 32B
- Price
- $0.10 / $0.30
- Speed
- 55 tok/s · 0.4s TTFT
- Modalities
- text → text
- License
- Apache 2.0
Fully open Ai2 reasoning model with public training data - transparent alternative to closed frontiers.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.