Pricing
- Input Tokens: $0.155 /M tokens
- Output Tokens: $0.62 /M tokens
- Cache Read: $0.0031 /M tokens
Input Modalities
- Text
- Vision
Output Modalities
- Text
Context length
- 1M tokens
Max output
- 393K tokens
Capabilities
- Thinking
- Streaming
- Tool calling
- Web search
- URL context
- Code interpreter
- Computer use
- File search
- Memory tool
- Structured outputs
- Structured decision
- Citations
- Prompt caching
- Background mode
- Server-side sessions
Providers
- Overall
- 66.37
- Response quality
- 50.83
- Feature support
- 62.24
- Response speed
- 76.57
- Service reliability
- 88.50
- Overall
- 56.48
- Response quality
- 36.70
- Feature support
- 75.71
- Response speed
- 36.55
- Service reliability
- 87.00
- Overall
- 68.72
- Response quality
- 53.97
- Feature support
- 85.33
- Response speed
- 52.50
- Service reliability
- 90.00
- Overall
- 73.38
- Response quality
- 52.13
- Feature support
- 80.54
- Response speed
- 85.01
- Service reliability
- 90.00
- Overall
- 73.76
- Response quality
- 52.59
- Feature support
- 84.60
- Response speed
- 80.99
- Service reliability
- 90.00
- Overall
- 64.10
- Response quality
- 50.42
- Feature support
- 60.16
- Response speed
- 68.55
- Service reliability
- 88.50
Dimension scores are D-layer composites (0.65 A + 0.20 B + 0.15 C); raw A-layer values are in each APB Profile.
Scoring methodology
Response quality quantifies task-level correctness under the specified adjudication protocol; feature support assesses conformance to API, tool-invocation and usage-reporting contracts; response speed characterizes latency under the evaluated load conditions; service reliability measures first-attempt completion within the observed request cohort.
Each dimension is a 0–100 composite of absolute performance, deviation from the reference route and evidence sufficiency. The overall score weights them 35%, 25%, 20% and 20%. These results are provisional and conditional on the sampled workload and observation protocol; they do not establish statistical superiority, a production ranking or a long-term SLA.
Performance for deepseek-v4.1-flash-1017
Uptime is the percentage of requests that succeeded over the past 72 hours. AIHubMix continuously monitors every provider and automatically retries with the next-best provider when one returns an error or responds too slowly; Latency is total round-trip time (lower is better); Throughput is how fast the model writes (tokens per second, higher is better).