Leaderboard

GLM 5.3 Flash (uncensored)

Two DGX Sparks

On two DGX Sparks it decodes 55.6 tok/s and reads a 48,000-token codebase at 1,802 tok/s. It gets 6 of 7 one-shot builds working. Ray tested it on 2026-10-18.

Machine
Two DGX Sparks: 2 × NVIDIA GB10, 2 × 128 GB unified memory, linked at 200 Gb/s
Engine
TensorFold (Mia’s AI Lab kit 1.8)
Weights
EXL3 4 bpw uncensored community build (neko-legends)
Recipe
DFlash2 drafts, tensor parallel over both Sparks
Made by
Z.ai
Size
MoE
Tested
2026-10-18
Asked who made it
“Is there something specific about my capabilities or development that you’d like to know more about?”
Overall
—
Builds
929
Knowledge
—
Speed
504

Scores out of 1,000. Overall is the mean of the other three; how each is scored.

Speed

Decode speed
55.6 tok/s
Decode, prose
47.2 tok/s
Decode, Snake
64.0 tok/s
First token
273 ms
Prefill, 48k prompt
1,802 tok/s
Tokens for the seven builds
201,832
Time for the seven builds
1h 03m

Decode speed is the median of a prose prompt and a Snake game, one request at a time. Prefill is how fast it reads a 48,000-token codebase before it answers.

Accuracy

SetCorrectScoreMinutes
MMLU-Pro, 140 questions, thinking low———
GSM8K, 100 problems, thinking low———
Snake by GLM 5.3 Flash (uncensored)
SnakeWorks
64.0 tok/s4m 00s15,331 tokens
Rocket launch by GLM 5.3 Flash (uncensored)
Rocket launchWorks
48.8 tok/s13m 01s38,112 tokens
Heptagon balls by GLM 5.3 Flash (uncensored)
Heptagon ballsWorks
63.8 tok/s56.1 s3,544 tokens
Solar system by GLM 5.3 Flash (uncensored)
Solar systemWorks
53.1 tok/s14m 43s46,934 tokens
Pelican on a bicycle by GLM 5.3 Flash (uncensored)
Pelican on a bicycleRuns with faults
58.7 tok/s1m 42s6,024 tokens
A complete, moving scene, but drawn as a small card in the middle of the page.
City fireworks by GLM 5.3 Flash (uncensored)
City fireworksWorks
54.6 tok/s9m 55s32,504 tokens
Gummy bears by GLM 5.3 Flash (uncensored)
Gummy bearsWorks
50.6 tok/s19m 33s59,383 tokens
Real gummy-bear shapes on a wooden table; about half of them land outside the bowl.

Run local models in Ownlight

Ownlight connects the model servers you already run, like Ollama, LM Studio, Unsloth Studio, llama-swap or Model Deck, and starts Claude Code or Codex on a local model beside your cloud agents. Local is part of Free.

Get OwnlightDownload for MacHow Local works

Every step

Decode speed, time to first token, prefill speed, output tokens and total time for each prompt in the suite.

TestDecode tok/sFirst tokenPrefill tok/sOutput tokensTotal timeRun
Prose (speed)47.2273 ms —3,7961m 20s811
Snake64.0615 ms —15,3314m 00s813
Read a 48k-token codebase49.223.4 s 1,80233530.2 s814
Who made you?52.1207 ms —2485.0 s815
Rocket launch48.8504 ms —38,11213m 01s816
Heptagon balls63.8500 ms —3,54456.1 s817
Solar system53.1477 ms —46,93414m 43s818
Pelican on a bicycle58.7277 ms —6,0241m 42s819
City fireworks54.6463 ms —32,5049m 55s820
Gummy bears50.6687 ms —59,38319m 33s821

More models on the bench

Updated 2026-10-20 16:39.