vs
0:4
overall capability wins (0 ties across 4 scored). Mercury 2 leads overall.
- Reasoning0:2
- Coding0:1
- Pricing0:3
- Specs0:1
Verdict
Mercury 2 leads coding (widest gap: +25.9 on SciCode); Mercury 2 offers ~3.9x larger context (128K vs 33K); DBRX Instruct ships open weights (DBRX).
- Mercury 2 leads coding (widest gap: +25.9 on SciCode)
- Mercury 2 offers ~3.9x larger context (128K vs 33K)
- DBRX Instruct ships open weights (DBRX)
- Mercury 2 is ~3.0x cheaper on a blended token basis than DBRX Instruct
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- DBRX Instruct
- $1.31
- Mercury 2
- $0.44
Mercury 2 is about 3.0x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Databricks
Inception Labs
Family
Other
Other
License
DBRX
Proprietary
Open weights
Yes
No
Release
Mar 27, 2024
Sep 1, 2026
Knowledge cutoff
2023-12
-
API / provider
Databricks Mosaic AI
Inception Labs
Modalities
text → text
text → text
Specs
Context window
33K
128K
Max output
8K
-
Parameters
132B (36B act.)
—
Pricing
Input $/1M
$0.75
$0.25
Output $/1M
$2.25
$0.75
Blended $/1M (3∶1)
$1.13
$0.38
Speed
tok/s
55
-
TTFT (s)
0.5
-
Reasoning
MMLU-Pro
39.7
—
GPQA Diamond
33.1
77.0
Humanity's Last Exam
2.9
17.1
AIME 2025
—
—
MATH-500
27.9
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
21.2
AA-LCR v1.1
—
43.7
CritPt
—
0.8
MMMU-Pro
—
—
IFBench
—
69.8
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
9.3
—
Terminal-Bench 2.1
—
27.3
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
11.8
37.7
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
1,167
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,195
—
| Metric | DBRX Instruct | Mercury 2 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Databricks | Inception Labs | — |
| Family | Other | Other | — |
| License | DBRX | Proprietary | — |
| Open weights | Yes | No | — |
| Release | Mar 27, 2024 | Sep 1, 2026 | — |
| Knowledge cutoff | 2023-12 | - | — |
| API / provider | Databricks Mosaic AI | Inception Labs | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 33K | 128K | B +95K |
| Max output | 8K | - | — |
| Parameters | 132B (36B act.) | — | — |
| Pricing | |||
| Input $/1M | $0.75 | $0.25 | B +$0.50 |
| Output $/1M | $2.25 | $0.75 | B +$1.50 |
| Blended $/1M (3∶1) | $1.13 | $0.38 | B +$0.75 |
| Speed | |||
| tok/s | 55 | - | — |
| TTFT (s) | 0.5 | - | — |
| Reasoning | |||
| MMLU-Pro | 39.7 | — | — |
| GPQA Diamond | 33.1 | 77.0 | B +43.9 pts |
| Humanity's Last Exam | 2.9 | 17.1 | B +14.2 pts |
| AIME 2025 | — | — | — |
| MATH-500 | 27.9 | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 21.2 | — |
| AA-LCR v1.1 | — | 43.7 | — |
| CritPt | — | 0.8 | — |
| MMMU-Pro | — | — | — |
| IFBench | — | 69.8 | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 9.3 | — | — |
| Terminal-Bench 2.1 | — | 27.3 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 11.8 | 37.7 | B +25.9 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | 1,167 | — |
| 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.
Mercury 2
- Context
- 128K
- Parameters
- —
- Price
- $0.25 / $0.75
- Speed
- —
- Modalities
- text → text
- License
- Proprietary
Inception Labs' diffusion-based (non-autoregressive) reasoning LLM, API-only with a 128K context window and high-throughput inference on Blackwell GPUs.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.