vs
0:5
overall capability wins (0 ties across 5 scored). K2 Horizon 375B A23B leads overall.
- Reasoning0:2
- Coding0:1
- Specs0:2
Verdict
K2 Horizon 375B A23B leads coding (widest gap: +31.1 on SciCode); K2 Horizon 375B A23B offers ~16.0x larger context (524K vs 33K).
- K2 Horizon 375B A23B leads coding (widest gap: +31.1 on SciCode)
- K2 Horizon 375B A23B offers ~16.0x larger context (524K vs 33K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- DBRX Instruct
- $1.31
- K2 Horizon 375B A23B
- -
Price per 1M tokens
Full comparison
Organization
Databricks
MBZUAI Institute of Foundation Models
Family
Other
Other
License
DBRX
Apache 2.0
Open weights
Yes
Yes
Release
Mar 27, 2024
Sep 3, 2026
Knowledge cutoff
2023-12
-
API / provider
Databricks Mosaic AI
No primary API price listed
Modalities
text → text
text → text
Specs
Context window
33K
524K
Max output
8K
-
Parameters
132B (36B act.)
375B (23B act.)
Pricing
Input $/1M
$0.75
—
Output $/1M
$2.25
—
Blended $/1M (3∶1)
$1.13
-
Speed
tok/s
55
-
TTFT (s)
0.5
-
Reasoning
MMLU-Pro
39.7
—
GPQA Diamond
33.1
87.3
Humanity's Last Exam
2.9
32.0
AIME 2025
—
—
MATH-500
27.9
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
18.2
AA-LCR v1.1
—
80.0
CritPt
—
4.6
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
9.3
—
Terminal-Bench 2.1
—
71.9
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
11.8
42.9
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,195
—
| Metric | DBRX Instruct | K2 Horizon 375B A23B | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Databricks | MBZUAI Institute of Foundation Models | — |
| Family | Other | Other | — |
| License | DBRX | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Mar 27, 2024 | Sep 3, 2026 | — |
| Knowledge cutoff | 2023-12 | - | — |
| API / provider | Databricks Mosaic AI | No primary API price listed | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 33K | 524K | B +492K |
| Max output | 8K | - | — |
| Parameters | 132B (36B act.) | 375B (23B act.) | B +243 |
| Pricing | |||
| Input $/1M | $0.75 | — | — |
| Output $/1M | $2.25 | — | — |
| Blended $/1M (3∶1) | $1.13 | - | — |
| Speed | |||
| tok/s | 55 | - | — |
| TTFT (s) | 0.5 | - | — |
| Reasoning | |||
| MMLU-Pro | 39.7 | — | — |
| GPQA Diamond | 33.1 | 87.3 | B +54.2 pts |
| Humanity's Last Exam | 2.9 | 32.0 | B +29.1 pts |
| AIME 2025 | — | — | — |
| MATH-500 | 27.9 | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 18.2 | — |
| AA-LCR v1.1 | — | 80.0 | — |
| CritPt | — | 4.6 | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 9.3 | — | — |
| Terminal-Bench 2.1 | — | 71.9 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 11.8 | 42.9 | B +31.1 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,195 | — | — |
DBRX Instruct
- Context
- 33K / 8K out
- Parameters
- 132B (36B act.)
- Price
- $0.75 / $2.25
- Speed
- 55 tok/s · 0.5s TTFT
- Modalities
- text → text
- License
- DBRX
Fine-grained 16-expert MoE (132B / 36B active) — strong open coding and SQL baseline from Databricks Mosaic.
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.