DeepSeek: DeepSeek V3.2 Exp
deepseek/deepseek-v3.2-exp
Cheapest provider
$0.27 / 1M
AtlasCloud
Fastest provider (p95)
55 tok/s
AtlasCloud
Intelligence (composite)
83.3
MMLU-Pro · HumanEval · math · GPQA
Per-provider performance
Latency / throughput / uptime measured across providers over the last 30 minutes of live traffic. Atlas’s router weighs these per call (with the eval-gate signal) when picking a variant for Standard and Batch tiers.
| Provider | Quant | Input $/1M | Output $/1M | Latency p50 / p95 | Throughput p50 / p95 | Uptime 30m | Success |
|---|---|---|---|---|---|---|---|
| AtlasCloud | q8· fp8 | $0.2700 | $0.4100 | 1262ms / 2704ms | 29 / 55 tok/s | 98.77% | 99.3% |
| Novita | q8· fp8 | $0.2700 | $0.4100 | 1369ms / 2666ms | 18 / 35 tok/s | 100.00% | 100.0% |
| SiliconFlow | q8· fp8 | $0.2700 | $0.4100 | 2126ms / 6448ms | 18 / 26 tok/s | 99.23% | 99.3% |
“—” means live telemetry hasn’t accumulated enough recent traffic for that endpoint. “undisclosed” means the provider serves the model but doesn’t expose the quantization label (typically running fp8 / int8 internally).
Intelligence breakdown
Composite score is a weighted average of public benchmarks (30% MMLU-Pro, 25% code pass@1, 25% math, 20% GPQA). Numbers come from model cards and the Artificial Analysis intelligence harness; missing components are renormalised over what’s present.
MMLU-Pro
85.0
broad reasoning
Code
79.0
pass@1 (HumanEval / LiveCodeBench)
AIME 2025
88.0
math accuracy
GPQA Diamond
80.0
hard reasoning
Source: Artificial Analysis (AA-aligned) via MiniMax-M2 table — DeepSeek-V3.2 (MMLU-Pro 85, GPQA-D 80, LiveCodeBench 79, AIME25 88); github.com/MiniMax-AI/MiniMax-M2
How Atlas routes DeepSeek: DeepSeek V3.2 Exp
- Realtime tier — direct passthrough at the upstream’s native precision. Best for hard latency / quality guarantees.
- Standard tier — Atlas picks the cheapest provider variant whose quantization has stayed green on your operation’s eval gates. For DeepSeek: DeepSeek V3.2 Exp that’s currently AtlasCloud at $0.27/1M.
- Batch tier — async, biggest discount. Roadmapped to use provider batch APIs (OpenAI / Anthropic) where available and queued spot capacity for open-weight workloads.