vs
6:5
overall capability wins (0 ties across 11 scored). Mercury 2 leads overall.
- Reasoning4:2
- Coding1:2
- Tool use1:0
- Composite indices2:0
- Specs0:1
Verdict
Trinity Large Thinking leads coding (widest gap: +70.0 on WebDev Elo); Mercury 2 edges reasoning & knowledge; Trinity Large Thinking offers ~2.0x larger context (262K vs 128K).
- Trinity Large Thinking leads coding (widest gap: +70.0 on WebDev Elo)
- Mercury 2 edges reasoning & knowledge
- Trinity Large Thinking offers ~2.0x larger context (262K vs 128K)
- Trinity Large Thinking ships open weights (Apache 2.0)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Mercury 2
- $0.44
- Trinity Large Thinking
- -
Price per 1M tokens
Full comparison
Organization
Inception Labs
Arcee AI
Family
Other
Other
License
Proprietary
Apache 2.0
Open weights
No
Yes
Release
Sep 1, 2026
Jan 27, 2026
Knowledge cutoff
-
-
API / provider
Inception Labs
No primary API price listed
Modalities
text → text
text → text
Specs
Context window
128K
262K
Max output
-
-
Parameters
—
400B (13B act.)
Pricing
Input $/1M
$0.25
—
Output $/1M
$0.75
—
Blended $/1M (3∶1)
$0.38
-
Speed
tok/s
-
-
TTFT (s)
-
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
77.0
75.2
Humanity's Last Exam
17.1
15.8
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
21.2
22.5
AA-LCR v1.1
43.7
38.0
CritPt
0.8
0.9
MMMU-Pro
—
—
IFBench
69.8
56.3
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
27.3
20.6
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
37.7
40.6
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
1,167
1,237
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
1,369
| Metric | Mercury 2 | Trinity Large Thinking | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Inception Labs | Arcee AI | — |
| Family | Other | Other | — |
| License | Proprietary | Apache 2.0 | — |
| Open weights | No | Yes | — |
| Release | Sep 1, 2026 | Jan 27, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Inception Labs | No primary API price listed | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 128K | 262K | B +134K |
| Max output | - | - | — |
| Parameters | — | 400B (13B act.) | — |
| Pricing | |||
| Input $/1M | $0.25 | — | — |
| Output $/1M | $0.75 | — | — |
| Blended $/1M (3∶1) | $0.38 | - | — |
| Speed | |||
| tok/s | - | - | — |
| TTFT (s) | - | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 77.0 | 75.2 | A +1.8 pts |
| Humanity's Last Exam | 17.1 | 15.8 | A +1.3 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 21.2 | 22.5 | B +1.3 pts |
| AA-LCR v1.1 | 43.7 | 38.0 | A +5.7 pts |
| CritPt | 0.8 | 0.9 | B +0.1 pts |
| MMMU-Pro | — | — | — |
| IFBench | 69.8 | 56.3 | A +13.5 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 27.3 | 20.6 | A +6.7 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 37.7 | 40.6 | B +2.9 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | 1,167 | 1,237 | B +70 Elo |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | 1,369 | — |
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.
Trinity Large Thinking
- Context
- 262K
- Parameters
- 400B (13B act.)
- Price
- —
- Speed
- —
- Modalities
- text → text
- License
- Apache 2.0
Arcee AI's reasoning-tuned Trinity Large (400B total / 13B active MoE, 4-of-256 experts), Apache 2.0 open weights, larger sibling to Trinity Mini; released Jan 27, 2026.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.