Skip to main content
ModelScale

Atria Dawn Preview

Shanghai Artificial Intelligence Laboratory · Open Weight · bench-align-v5

Self-hosted
Canonical idatria-dawn-preview
Overall scoreUnavailable
Context window256K tokens
Release date2026-09-14
Access typeOpen Weight
Blended $/1MUnavailable75% input / 25% output

Capability shape

Seven axes from the ranking source. A missing axis is drawn as a gap.

Capability evidence

Agentic82.8
Coding44.7
KnowledgeUnavailable
ReasoningUnavailable
Multimodal & GroundedUnavailable
Instruction FollowingUnavailable
MathUnavailable

Runtime service evidence

Measured values with the date they were observed. Nothing is inferred from a sibling model or a provider claim.

Time to first token: latency to the first answer chunk (Artificial Analysis, via BenchLM). Reasoning models include thinking time, so values can run to tens or hundreds of seconds.

Evidence key: ObservedUnavailable

Runtime measurements
MeasurementValueObservedLast goodEvidence
Time to first tokenUnavailableUnavailableUnavailable
ThroughputUnavailableUnavailableUnavailable

Regional or per-endpoint measurements appear only when the API supplies them; none are modelled here.

Endpoint and price matrix

Every published price component, including cache reads and writes.

Price components
ComponentUSDEvidence
Input / 1M tokensUnavailable
Output / 1M tokensUnavailable
Cache read / 1M tokensUnavailable
Cache write / 1M tokensUnavailable
Blended / 1M (75% input / 25% output)Unavailable
Self-hosted listingShanghai Artificial Intelligence Laboratory publishes the full-precision and FP8 Atria Dawn Preview checkpoints under MIT for self-hosting. The pricing catalog represents the open-weight row as self-host/free-per-token before infrastructure costs. Atria also offers the exact `Atria-Dawn-Preview` model through its hosted API, but its public site and API documentation do not publish a first-party token rate.No hosted token rate was published for this model, so its per-token price is unavailable rather than zero.

Workload-aware monthly cost example

10 conversations per day × 8 messages × 22 active days, 1200 input and 400 output tokens per message, no cache. This uses the same calculator as the cost simulator, so an unavailable applicable rate makes the example unavailable too.

Modelled monthly costUnavailableThe applicable input rate is unavailable.
Modelled tokensUnavailable
Open the simulatorChange this workload

Benchmark record

13 matched benchmark rows with their published value, unit, and provenance.

Coding

Coding benchmarks
BenchmarkValueTasksDifficultyProvenance
Terminal-Bench 2.1
Terminal-Bench 2.1 (provider run)
78.3Terminal-based software-agent tasksProfessional software engineeringDeepSeek V4 Flash 0731 update
SWE-bench Pro
SWE-bench Pro
59.61,865 repository problemsLong-horizon professional engineeringSWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?

Agentic

Agentic benchmarks
BenchmarkValueTasksDifficultyProvenance
Terminal-Bench 2.1
Terminal-Bench 2.1 (provider run)
78.3Terminal-based software-agent tasksProfessional software engineeringDeepSeek V4 Flash 0731 update
BrowseComp
BrowseComp
92.5Research questions requiring browsingHard web researchBrowseComp
GDPval-AA
GDPval-AA
1583Agentic real-world work tasksProfessional agentic workflowsDeepSeek-V4 Technical Report
CyberGym
CyberGym
86.51,507 vulnerability analysis instancesReal-world cybersecurityCyberGym: Evaluating AI Agents' Real-World Cybersecurity Capabilities at Scale
JobBench
JobBench
50.3130 tasks across 35 occupationsProfessional multi-source workflowsJobBench: Aligning Agent Work With Human Will
AutomationBench
AutomationBench
53.8600 public automation tasksLong-horizon automationKimi K3: Open Frontier Intelligence
DeepSearchQA
DeepSearchQA
96.0Agentic browsing and list-answer questionsAgentic web researchMuse Spark Eval Methodology
BFCL v4
Berkeley Function Calling Leaderboard v4
77.0Function-calling tasksAdvanced tool useTrinity-Large-Thinking: Scaling an Open Source Frontier Agent
MLE-Bench Lite
MLE-Bench Lite
86.2Low-resource ML competitionsAgentic machine learningMiniMax M2.7: Early Echoes of Self-Evolution
τ³-bench results
τ³-Bench Tool-Agent-User Evaluation
41.2Corrected customer-service tasks plus knowledge and voice evaluation modesLong-horizon, multimodal, and knowledge-aware tool useOfficial τ³-bench repository and release notes
WideResearch
WideResearch
81.9Open-ended research tasksBroad research-agent workflowsQwen3.6 launch benchmarks

Lifecycle and limitations log

Lifecycle events the source associates with this model, plus what this profile does not claim.

No lifecycle event references this model.

That is not evidence the model has no lifecycle plan — only that this source published none.

What this profile does not claimValues are reproduced exactly as their sources published them, in the units those sources declared; none are converted, interpolated, or averaged across providers. Any field marked unavailable was attempted and not returned — the attempt timestamp is in each badge. Last attempted fetch for this model’s score: 2026-09-17 17:17 UTC.