Running DeepSeek V4.1 Flash on MI325X
Quick answer
DeepSeek V4.1 Flash sustains 17,745 tokens/s per GPU on MI325X at 50 tokens/s per user, which works out to $0.017 per million tokens at hyperscaler pricing, served by vLLM. Fastest measured TTFT: 0.4 ms; fastest TPOT: 0.0 ms (each the best across all configs, not one run).
Benchmarked configs
18
Serving engines
vLLM
Precisions
fp4
Run dates
2026-09-22 β 2026-09-22
Throughput at every interactivity target
Serving is a trade-off: push more concurrent users through a GPU and each user's tokens arrive slower. The ladder below reads the measured frontier at each per-user speed target on the AgentX agentic coding workload, using the best engine and precision at that point.
| Per-user target | Tokens/s per GPU | $ / 1M tokens | Engine | Precision |
|---|---|---|---|---|
| 30 tok/s | 24,472 | $0.012 | vLLM | fp4 |
| 50 tok/s | 17,745 | $0.017 | vLLM | fp4 |
| 75 tok/s | 13,945 | $0.022 | vLLM | fp4 |
| 100 tok/s | 11,287 | $0.027 | vLLM | fp4 |
| 150 tok/s | 4,673 | $0.065 | vLLM | fp4 |
| 200 tok/s | 3,435 | $0.089 | vLLM | fp4 |
What serving actually costs
Converting the 50 tokens/s per user operating point to $ per million total tokens across rental pricing tiers from the SemiAnalysis AI Cloud TCO model.
| Pricing tier | $ / GPU / hr | $ / 1M tokens |
|---|---|---|
| Owning at Large Hyperscaler Volume | $1.10 | $0.017 |
| Retail | $1.60 | $0.025 |
Frequently asked questions
- How fast is DeepSeek V4.1 Flash on MI325X?
- At an interactivity target of 50 tokens/s per user on the AgentX agentic coding workload, MI325X sustains 17,745 tokens/s per GPU serving DeepSeek V4.1 Flash with vLLM in FP4. Peak measured throughput across all configs is 32,095 tokens/s per GPU.
- How much does it cost to serve DeepSeek V4.1 Flash on MI325X?
- $0.017 per million total tokens at large-hyperscaler-volume ownership $/GPU/hr pricing, at 50 tokens/s per user. The retail rental tier is tabulated above; slower interactivity targets lower the cost further.
- Which serving engines run DeepSeek V4.1 Flash on MI325X?
- The runs behind this page used vLLM in FP4. Engines are rebuilt and re-benchmarked continuously, so the best config can change between visits.
- How are these DeepSeek V4.1 Flash numbers measured?
- Every number is measured on real MI325X hardware by the InferenceX fleet, sweeping concurrency on the AgentX agentic coding workload to trace the throughput-versus-interactivity frontier; the newest run landed on 2026-09-22. The same derivation powers the InferenceX overview leaderboard.