open god mode ▮

Z.ai: GLM 5.3 Flash glm-5.3-flash

zai-org · zai-org radar-draft

#3 overall

params
321.3B
arch
moe
context
1024k
license
mit open weights
released
Aug 2026
train compute
3.2e+24 FLOP
downloads/30d
6.4M

compare ↔ hugging face ↗

# architecture

attn shared MHA ×64 router top-8 of 288 +1 shared expert ×288 expert out ×45 layers ctx 1,048,576 321.3B pool
layers
45
d_model
4096
heads
64q / 64kv
head dim
64
experts
top-8 of 288 + 1 shared
vocab
154880
family
glm5_next_text

signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.

# why ranked

Overall: #3 — score 97.2 ● low — a single signal; treat with caution

signalweightinput (0–100)
aa_intelligence 0.70 95.9
bench_composite 0.30 100.0

benchmark panel evidence:

complete panelscore (0–100)
aa-coding-v1100.0

Coding: provisional — score 94.1 ● low — a single signal; treat with caution

provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.

signalweightinput (0–100)
aa_coding 0.30 94.1
aider_polyglot 0.30 —
swe_bench_verified 0.40 —

missing signals are dropped and the remaining weights renormalized — never imputed.

full methodology

# trend

OGM score · last 45 days
OGM score over 45 days
overall rank (up is better)
Overall rank over 45 days

# benchmarks

benchmarkscoresourcedate
AA Coding Index 71.5 Artificial Analysis —
AA Intelligence Index 41.8 Artificial Analysis —
Agents-last-exam 0.3 / 1 LLM Stats —
Artificial-analysis 0.6 / 1 LLM Stats —
Automationbench 0.5 / 1 LLM Stats —
Babyvision 0.5 / 1 LLM Stats —
Chartography 0.8 / 1 LLM Stats —
Charxiv-r 0.9 / 1 LLM Stats —
Gdpval-aa 1773.0 / 3000 LLM Stats —
LMArena Elo 1473.0 LMArena Oct 2026
LMArena Elo (Code) 1616.0 LMArena Oct 2026
LMArena Elo (Vision) 1278.0 LMArena Oct 2026

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime✓
Featherless — ——— via hfrouter — —
Relaceunknown 1024k $0.04$0.08$0.03 via openrouter — 100.0%
OpenInferencefp4 1024k $0.08$0.01$0.02 via openrouter — 100.0%
StreamLakefp8 1000k $0.04$0.14$0.01 via openrouter — 100.0%
Wafernvfp4 1024k $0.05$0.20$0.05 via openrouter — 100.0%
Darkbloomfp4 256k $0.07$0.18$0.03 via openrouter — 96.8%
GMICloudfp8 1024k $0.07$0.25$0.01 via openrouter — 95.9%
Sail Researchfp4 1024k $0.08$0.25$0.03 via openrouter — 100.0%
Sail Researchfp4 1024k $0.08$0.25$0.03 via openrouter — 100.0%
Vultr 1024k $0.10$0.35— via modelsdev — —
Sail 1024k $0.11$0.35— via litellm — —
Aihubmix 1024k $0.11$0.39— via litellm — —
Near AIfp8 1024k $0.08$0.50$0.01 via openrouter — 100.0%
InferenceNetfp4 1024k $0.12$0.39$0.03 via openrouter — 100.0%
InferenceNetfp4 1024k $0.12$0.39$0.03 via openrouter — 100.0%
DigitalOceanunknown 1024k $0.13$0.44$0.03 via openrouter — 100.0%
NextBitfp4 1024k $0.12$0.50$0.05 via openrouter — 100.0%
Makoraunknown 1024k $0.14$0.47$0.02 via openrouter — 100.0%
AtlasCloudfp8 1024k $0.15$0.50— atlascloud — 98.4% ✓
BaseTenfp8 1024k $0.15$0.50$0.03 via openrouter — —
BaseTenfp8 1024k $0.15$0.50$0.03 via openrouter — 100.0%
Cloudflare 1024k $0.15$0.50$0.03 via modelsdev — —
Cohereunknown 1024k $0.15$0.50$0.03 via openrouter — 100.0%
CoreWeavenvfp4 1024k $0.15$0.50$0.02 via openrouter — 99.7%
Crusoefp4 1024k $0.15$0.50$0.03 via openrouter — 100.0%
DeepInfrafp4 1024k $0.15$0.50— deepinfra — 100.0% ✓
Fireworksunknown 1024k $0.15$0.50$0.03 via openrouter — 99.9%
Friendliunknown 1024k $0.15$0.50$0.03 via openrouter — 100.0%
Friendliai 1024k $0.15$0.50— via litellm — —
Mistralunknown 1024k $0.15$0.50$0.03 via openrouter — 99.7%
Modalnvfp4 1024k $0.15$0.50$0.03 via openrouter — —
Nebiusfp4 1000k $0.15$0.50$0.03 openrouter — 90.8% ✓
Novitafp8 1024k $0.15$0.50— novita — 99.9% ✓
Opencode Go 1000k $0.15$0.50— via litellm — —
Parasailfp4 1024k $0.15$0.50$0.03 via openrouter — 100.0%
Phalanvfp4 1024k $0.15$0.50— phala — 99.8% ✓
PrimeIntellectunknown 1024k $0.15$0.50$0.03 via openrouter — 100.0%
SiliconFlowfp8 1024k $0.15$0.50$0.03 openrouter — 99.9% ✓
Tencent 1000k $0.15$0.50— via litellm — —
Togetherunknown 1048k $0.15$0.50$0.03 via openrouter — 99.9%
Veniceunknown 1024k $0.15$0.50— venice — 100.0% ✓
Wandb 1049k $0.15$0.50— via litellm — —
Z.AIfp8 1024k $0.15$0.50$0.03 via openrouter — 99.9%
Inceptronfp8 1024k $0.15$0.55$0.05 via openrouter — 100.0%
Morphfp8 1024k $0.19$0.65$0.05 via openrouter — 100.0%
Inferenceapis 1048k $0.20$0.66— via litellm — —
Fireworksunknown 1024k $0.23$0.75$0.04 via openrouter — 98.8%
Rekaunknown 1024k $0.15$2.40$0.08 via openrouter — 100.0%

sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price

source aliases
arena
glm-5.3-flash
epoch
GLM-5.3-Flash
llmstats
glm-5.3-flash
openrouter
z-ai/glm-5.3-flash