vs
1:5
overall capability wins (0 ties across 6 scored). K2 Horizon 375B A23B leads overall.
- Reasoning1:2
- Coding0:1
- Specs0:2
Verdict
K2 Horizon 375B A23B leads coding (widest gap: +4.8 on SciCode); K2 Horizon 375B A23B offers ~2.0x larger context (524K vs 262K).
- K2 Horizon 375B A23B leads coding (widest gap: +4.8 on SciCode)
- K2 Horizon 375B A23B offers ~2.0x larger context (524K vs 262K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Agnes-3.0-Flash Preview
- -
- K2 Horizon 375B A23B
- -
Price per 1M tokens
Full comparison
Organization
Agnes AI
MBZUAI Institute of Foundation Models
Family
Other
Other
License
Apache 2.0
Apache 2.0
Open weights
Yes
Yes
Release
Sep 13, 2026
Sep 3, 2026
Knowledge cutoff
-
-
API / provider
No primary API price listed
No primary API price listed
Modalities
text, image, video → text
text → text
Specs
Context window
262K
524K
Max output
-
-
Parameters
33B
375B (23B act.)
Pricing
Input $/1M
—
—
Output $/1M
—
—
Blended $/1M (3∶1)
-
-
Speed
tok/s
-
-
TTFT (s)
-
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
85.0
87.3
Humanity's Last Exam
—
32.0
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
23.0
18.2
AA-LCR v1.1
68.3
80.0
CritPt
—
4.6
MMMU-Pro
—
—
IFBench
74.2
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
—
71.9
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
38.1
42.9
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Agnes-3.0-Flash Preview | K2 Horizon 375B A23B | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Agnes AI | MBZUAI Institute of Foundation Models | — |
| Family | Other | Other | — |
| License | Apache 2.0 | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Sep 13, 2026 | Sep 3, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | No primary API price listed | No primary API price listed | — |
| Modalities | text, image, video → text | text → text | — |
| Specs | |||
| Context window | 262K | 524K | B +262K |
| Max output | - | - | — |
| Parameters | 33B | 375B (23B act.) | B +342 |
| Pricing | |||
| Input $/1M | — | — | — |
| Output $/1M | — | — | — |
| Blended $/1M (3∶1) | - | - | — |
| Speed | |||
| tok/s | - | - | — |
| TTFT (s) | - | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 85.0 | 87.3 | B +2.3 pts |
| Humanity's Last Exam | — | 32.0 | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 23.0 | 18.2 | A +4.8 pts |
| AA-LCR v1.1 | 68.3 | 80.0 | B +11.7 pts |
| CritPt | — | 4.6 | — |
| MMMU-Pro | — | — | — |
| IFBench | 74.2 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | — | 71.9 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 38.1 | 42.9 | B +4.8 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Agnes-3.0-Flash Preview
- Context
- 262K
- Parameters
- 33B
- Price
- —
- Speed
- —
- Modalities
- text, image, video → text
- License
- Apache 2.0
Agnes AI's (Sapiens AI, Singapore) first open-weight release: a 33B-parameter hybrid-attention model (54 gated delta-rule recurrent layers alternating with 18 global-attention layers, 3:1) under Apache 2.0, with a 262K-token context window and text/image/video input. This open-weight Preview checkpoint is distinct from Agnes AI's production "Agnes 3.0 Flash" API model, which runs a different checkpoint and is not open-weight; the two should not be conflated.
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.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.