GPT-OSS 20B gpt-oss-20b

openai · gpt-oss reasoning efficient

unranked — no quality signals yet

params
21B (A3.6B)
arch
moe
context
128k
license
apache-2.0 open weights
released
Aug 2025
reasoning
yes
train compute
5.5e+23 FLOP
downloads/30d
6.6M

# architecture

attn shared · reasoning GQA 64q/8kv router top-4 of 32 expert ×32 expert out ×24 layers ctx 131,072 21B pool · A3.6B/token
layers
24
d_model
2880
heads
64q / 8kv
head dim
64
experts
top-4 of 32
vocab
201088
family
gpt_oss

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: provisional — score 32.0 ● 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_intelligence 0.70 32.0
bench_composite 0.30

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

Coding: provisional — score 19.3 ● 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 19.3
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 55 days
OGM score over 55 days

methodology changed during this history window; score movement across that boundary is not model movement. See methodology v5.

# benchmarks

benchmarkscoresourcedate
AA Coding Index 20.7 Artificial Analysis
AA Intelligence Index 9.1 Artificial Analysis
AA Math Index 89.3 Artificial Analysis
Codeforces 0.7 / 3000 LLM Stats

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime
Featherless via hfrouter
AkashMLfp4 128k $0.02$0.10 via openrouter 99.5%
Darkbloomfp8 128k $0.02$0.10 via openrouter 99.8%
CoreWeavefp4 128k $0.03$0.13$0.03 via openrouter 99.9%
Wandb 128k $0.03$0.13 via litellm
DeepInfrabf16 128k $0.03$0.14 deepinfra 100.0%
Parasailfp4 128k $0.03$0.15$0.02 via openrouter 99.9%
Amazon Bedrockunknown 128k $0.03$0.15 bedrock 100.0%
Amazon Bedrockunknown 128k $0.03$0.15 bedrock 99.9%
Novitafp4 128k $0.04$0.15 novita 99.7%
Phalaunknown 128k $0.04$0.15 phala 95.5%
SiliconFlowfp8 128k $0.04$0.18 openrouter 98.7%
OVHcloud 128k $0.05$0.18 ovhcloud 45
Venicenot-available 125k $0.05$0.19 venice
Nscale 128k $0.05$0.20 via hfrouter 155
Togetherunknown 128k $0.05$0.20 via openrouter 99.3%
Googleunknown 128k $0.07$0.25 via openrouter 99.7%
Tensormesh 128k $0.07$0.28 via litellm
Fireworksunknown 128k $0.07$0.30$0.04 via openrouter
Bedrock Mantle 128k $0.07$0.30 via litellm
Groqunknown 128k $0.07$0.30 groq 99.6%
Replicate $0.09$0.36 via litellm
Cloudflare 125k $0.20$0.30 via modelsdev

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
aa
gpt-oss-20b
arena
gpt-oss-20b
openrouter
openai/gpt-oss-20b