vs
2:0
overall capability wins (1 ties across 3 scored). Atria Dawn Preview leads overall.
- Tool use1:0
- Specs1:0
Verdict
Atria Dawn Preview and Hy3 are closely matched on the overlapping metrics we have.
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Atria Dawn Preview
- -
- Hy3
- $0.26
Price per 1M tokens
Full comparison
Organization
Shanghai AI Lab
Tencent
Family
Other
Other
License
MIT
Apache 2.0
Open weights
Yes
Yes
Release
Sep 11, 2026
Jul 6, 2026
Knowledge cutoff
-
-
API / provider
No primary API price listed
Tencent Cloud TokenHub
Modalities
text → text
text → text
Specs
Context window
256K
256K
Max output
66K
-
Parameters
744B
295B (21B act.)
Pricing
Input $/1M
—
$0.13
Output $/1M
—
$0.53
Blended $/1M (3∶1)
-
$0.23
Speed
tok/s
-
-
TTFT (s)
-
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
—
89.7
Humanity's Last Exam
—
33.5
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
32.0
AA-LCR v1.1
—
79.0
CritPt
—
4.9
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
59.6
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
—
64.4
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
—
48.6
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
1,513
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
1,456
| Metric | Atria Dawn Preview | Hy3 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Shanghai AI Lab | Tencent | — |
| Family | Other | Other | — |
| License | MIT | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Sep 11, 2026 | Jul 6, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | No primary API price listed | Tencent Cloud TokenHub | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 256K | 256K | tie |
| Max output | 66K | - | — |
| Parameters | 744B | 295B (21B act.) | A +449 |
| Pricing | |||
| Input $/1M | — | $0.13 | — |
| Output $/1M | — | $0.53 | — |
| Blended $/1M (3∶1) | - | $0.23 | — |
| Speed | |||
| tok/s | - | - | — |
| TTFT (s) | - | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | — | 89.7 | — |
| Humanity's Last Exam | — | 33.5 | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 32.0 | — |
| AA-LCR v1.1 | — | 79.0 | — |
| CritPt | — | 4.9 | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | 59.6 | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | — | 64.4 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | — | 48.6 | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | 1,513 | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | 1,456 | — |
Atria Dawn Preview
- Context
- 256K / 66K out
- Parameters
- 744B
- Price
- —
- Speed
- —
- Modalities
- text → text
- License
- MIT
Shanghai AI Laboratory's open-weight agentic MoE (744B total, MIT license, 256K context), built on a GLM-5.2 base and trained by the cross-institutional ATRIA initiative (with Fudan NLP Lab) for long-horizon research and engineering workflows: problem analysis, tool use, code, experiments, and failure recovery. Benchmark scores are from the model's own technical report and have not yet been independently verified by a third-party evaluator.
Hy3
- Context
- 256K
- Parameters
- 295B (21B act.)
- Price
- $0.13 / $0.53
- Speed
- —
- Modalities
- text → text
- License
- Apache 2.0
Tencent's July 2026 general-availability Hunyuan 3 release (295B/21B-active MoE, Apache 2.0) succeeding the April 2026 Hy3 preview with expanded post-training on feedback from 50+ products.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.