Granite 3.3 8B Instruct granite-3.3-8b-instruct

ibm-granite · ibm-granite

unranked — no quality signals yet

params
8.2B
arch
dense
license
apache-2.0 open weights

# architecture

attn shared GQA 32q/8kv feed-forward (dense) every parameter, every token out ×40 layers 8.2B — no routing
layers
40
d_model
4096
heads
32q / 8kv
head dim
128
vocab
49159
family
granite

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.

# trend

OGM score · last 93 days
OGM score over 93 days
overall rank (up is better)
Overall rank over 93 days

methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.

# benchmarks

benchmarkscoresourcedate
AIME 2024 0.8 / 1 LLM Stats —
Alpacaeval-2.0 0.6 / 1 LLM Stats —
Arena-hard 0.6 / 1 LLM Stats —
Attaq 0.9 / 1 LLM Stats —
Big-bench-hard 0.7 / 1 LLM Stats —

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime✓
Replicate — $0.03$0.25— via litellm — —

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
llmstats
granite-3.3-8b-instruct