Running GLM-5.3 on B300
Quick answer
GLM-5.3 sustains 34,534 tokens/s per GPU on B300 at 50 tokens/s per user, which works out to $0.018 per million tokens at hyperscaler pricing, served by Dynamo SGLang. Fastest measured TTFT: 0.2 ms; fastest TPOT: 0.0 ms (each the best across all configs, not one run).
Benchmarked configs
25
Serving engines
Dynamo SGLang, SGLang
Precisions
fp4, fp8
Run dates
2026-10-12 β 2026-10-14
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 | - | - | Dynamo SGLang | fp4 |
| 50 tok/s | 34,534 | $0.018 | Dynamo SGLang | fp4 |
| 75 tok/s | 24,345 | $0.026 | Dynamo SGLang | fp4 |
| 100 tok/s | 20,709 | $0.030 | Dynamo SGLang | fp4 |
| 150 tok/s | 14,310 | $0.044 | Dynamo SGLang | fp4 |
| 200 tok/s | 9,558 | $0.066 | Dynamo SGLang | 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 | $2.26 | $0.018 |
| Retail | $4.25 | $0.034 |
Frequently asked questions
- How fast is GLM-5.3 on B300?
- At an interactivity target of 50 tokens/s per user on the AgentX agentic coding workload, B300 sustains 34,534 tokens/s per GPU serving GLM-5.3 with Dynamo SGLang in FP4. Peak measured throughput across all configs is 42,047 tokens/s per GPU.
- How much does it cost to serve GLM-5.3 on B300?
- $0.018 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 GLM-5.3 on B300?
- The runs behind this page used Dynamo SGLang, SGLang in FP4, FP8 and multi-node serving. Engines are rebuilt and re-benchmarked continuously, so the best config can change between visits.
- How are these GLM-5.3 numbers measured?
- Every number is measured on real B300 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 2026-10-14. The same derivation powers the InferenceX overview leaderboard.