GLM-5.3 NVFP4
- Input
- —
- Output
- —
- Speed
- ~140tok/sest.
- Context
- 1Mtokens
Rankings
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs. No API sells this model, so this is its price on the chart.
- Hardware
- 8× H200 141GB
- NVFP4 weights
- Throughput
- ~16,000 tok/s
- many requests batched
- Price
- $0.25–2.08 /M tok
- busy → light use
- On your own machine
- Too large
- needs data-center GPUs
Assumptions (5)
- NVFP4 weights (744B params, 40B active per token) + 40% KV-cache headroom ≈ 586 GB
- 8× H200 141GB at $3.59–$7.91/GPU-hour on-demand (2026-09-24)
- ~16,000 output tok/s aggregate at batch 575 (bandwidth-bound); MoE compute scales with active params
- Blended 3:1 input:output; prefill ~93,508 tok/s
- Low = 75% utilization at the low GPU price; high = 20% at the high price
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench VerifiedEstimated
~#2 of 21 (estimated)~88.1%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#6 of 24 (estimated)~72.8%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #5Claude Opus 579.2%
- ~#6GLM-5.3 NVFP4Estimated72.8%
- #6Intern-S2-397B68.5%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#32 of 46 (estimated)~59.0%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #31Muse Spark Open60.0%
- ~#32GLM-5.3 NVFP4Estimated59.0%
- #32Qwen3.8 Flash-Next58.7%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#19 of 102 (estimated)~33.3%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #18DeepSeek V4.1 Flash 101733.8%
- ~#19GLM-5.3 NVFP4Estimated33.3%
- #19Ling 3.1 Flash33.3%
No published score — estimated from related benchmarks
LiveCodeBenchLab-reported
#17 of 2753.5%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #16Claude Haiku 4.561.5%
- #17GLM-5.3 NVFP453.5%
- #18Phi-4-reasoning-plus53.1%
huggingface.co · 2026-09-29
SciCodeLab-reported
#16 of 11858.0%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #15Muse Spark Open58.4%
- #16GLM-5.3 NVFP458.0%
- #17GLM-5.3 Flash NVFP457.7%
huggingface.co · 2026-09-29
AA Coding Agent IndexEstimated
~#10 of 26 (estimated)~52.8
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #9Gemini 4 Argon53.2
- ~#10GLM-5.3 NVFP4Estimated52.8
- #10Muse Wonder52.0
No published score — estimated from related benchmarks
Writing & chat
IFBenchLab-reported
#49 of 13966.0%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #47Nova 2.0 Omni66.2%
- #49GLM-5.3 NVFP466.0%
- #49Olmo 3.1 32B Think66.0%
huggingface.co · 2026-09-29
LMArena Text, Non-English (Elo)Estimated
~#18 of 57 (estimated)~1453 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #17GPT-5.6 Terra1456
- ~#18GLM-5.3 NVFP4Estimated1453
- #18DeepSeek V4 Pro (0813)1448
No published score — estimated from related benchmarks
LMArena Text (Elo)Estimated
~#17 of 57 (estimated)~1467 Elo
- #1Gemini 4 Argon1525
- #2Claude Opus 5.51507
- #16DeepSeek V4.1 Flash1474
- ~#17GLM-5.3 NVFP4Estimated1467
- #17GPT-5.6 Terra1466
No published score — estimated from related benchmarks
EQ-Bench Creative Writing v3 (Elo)Estimated
~#8 of 26 (estimated)~2015 Elo
- #1GPT-6 Astra2173
- #2Claude Fable 5.12162
- #7Claude Opus 5.52050
- ~#8GLM-5.3 NVFP4Estimated2015
- #8Grok 4.72007
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondLab-reported
#15 of 19292.7%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #13Qwen3.8 Max (0902)92.8%
- #15GLM-5.3 NVFP492.7%
- #16GPT-5.6 Terra92.5%
huggingface.co · 2026-09-29
Humanity's Last ExamEstimated
~#57 of 199 (estimated)~30.3%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #56Nemotron 3.5 Super VL30.7%
- ~#57GLM-5.3 NVFP4Estimated30.3%
- #57A.X-K229.6%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#38 of 103 (estimated)~32.7
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #37Qwen3.8 27B33.7
- ~#38GLM-5.3 NVFP4Estimated32.7
- #38K2 Horizon 375B A23B30.5
No published score — estimated from related benchmarks