vs
4:2
overall capability wins (0 ties across 6 scored). Mercury 2 leads overall.
- Reasoning4:1
- Pricing0:3
- Specs0:1
Verdict
Mercury 2 edges reasoning & knowledge; Seed-OSS-36B-Instruct offers ~4.0x larger context (512K vs 128K); Seed-OSS-36B-Instruct ships open weights (Apache 2.0).
- Mercury 2 edges reasoning & knowledge
- Seed-OSS-36B-Instruct offers ~4.0x larger context (512K vs 128K)
- Seed-OSS-36B-Instruct ships open weights (Apache 2.0)
- Seed-OSS-36B-Instruct is ~1.3x cheaper on a blended token basis than Mercury 2
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Mercury 2
- $0.44
- Seed-OSS-36B-Instruct
- $0.35
Seed-OSS-36B-Instruct is about 1.2x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Inception Labs
ByteDance Seed
Family
Other
Other
License
Proprietary
Apache 2.0
Open weights
No
Yes
Release
Sep 1, 2026
Aug 20, 2025
Knowledge cutoff
-
-
API / provider
Inception Labs
SiliconFlow
Modalities
text → text
text → text
Specs
Context window
128K
512K
Max output
-
-
Parameters
—
36B
Pricing
Input $/1M
$0.25
$0.21
Output $/1M
$0.75
$0.57
Blended $/1M (3∶1)
$0.38
$0.30
Speed
tok/s
-
-
TTFT (s)
-
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
77.0
72.6
Humanity's Last Exam
17.1
9.9
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
21.2
18.2
AA-LCR v1.1
43.7
61.3
CritPt
0.8
—
MMMU-Pro
—
—
IFBench
69.8
41.9
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
27.3
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
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
—
—
| Metric | Mercury 2 | Seed-OSS-36B-Instruct | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Inception Labs | ByteDance Seed | — |
| Family | Other | Other | — |
| License | Proprietary | Apache 2.0 | — |
| Open weights | No | Yes | — |
| Release | Sep 1, 2026 | Aug 20, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Inception Labs | SiliconFlow | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 128K | 512K | B +384K |
| Max output | - | - | — |
| Parameters | — | 36B | — |
| Pricing | |||
| Input $/1M | $0.25 | $0.21 | B +$0.04 |
| Output $/1M | $0.75 | $0.57 | B +$0.18 |
| Blended $/1M (3∶1) | $0.38 | $0.30 | B +$0.08 |
| Speed | |||
| tok/s | - | - | — |
| TTFT (s) | - | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 77.0 | 72.6 | A +4.4 pts |
| Humanity's Last Exam | 17.1 | 9.9 | A +7.2 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 21.2 | 18.2 | A +3.0 pts |
| AA-LCR v1.1 | 43.7 | 61.3 | B +17.6 pts |
| CritPt | 0.8 | — | — |
| MMMU-Pro | — | — | — |
| IFBench | 69.8 | 41.9 | A +27.9 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 27.3 | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 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 | — | — | — |
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.
Seed-OSS-36B-Instruct
- Context
- 512K
- Parameters
- 36B
- Price
- $0.21 / $0.57
- Speed
- —
- Modalities
- text → text
- License
- Apache 2.0
ByteDance Seed's open 36B dense model with a 512K native context window and a user-controllable reasoning 'thinking budget'.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.