modelgrep

AllenAI: Olmo 3 32B Think vs Meta: Llama 3.1 8B Instruct

AAllenAI: Olmo 3 32B Thinkallenai/olmo-3-32b-thinkVSMMeta: Llama 3.1 8B Instructmeta-llama/llama-3.1-8b-instruct

Meta: Llama 3.1 8B Instruct wins on more metrics (5 of 8), but the right pick depends on what you optimize for — see the breakdown below.

MetricAllenAI: Olmo 3 32B ThinkMeta: Llama 3.1 8B Instruct
Intelligence Index12.1✓11.8
Coding Index10.5✓4.9
GPQA Diamond61%✓26%
Design Arena Elo——
Speed (tokens/sec)—161✓
Latency—146ms✓
Input price /M$0.150$0.020✓
Output price /M$0.500$0.030✓
Context window66K131K✓
CapabilitiesReasoningJSONToolsJSON
AllenAI: Olmo 3 32B Think details →Meta: Llama 3.1 8B Instruct details →