modelgrep

IBM: Granite 4.0 Micro vs Meta: Llama 3.2 1B Instruct

IBM: Granite 4.0 Micro wins on more metrics (5 of 6), but the right pick depends on what you optimize for — see the breakdown below.

MetricIBM: Granite 4.0 MicroMeta: Llama 3.2 1B Instruct
Intelligence Index7.76.3
Coding Index5.00.6
GPQA Diamond34%20%
Design Arena Elo
Speed (tokens/sec)
Latency
Input price /M$0.017$0.027
Output price /M$0.112$0.201
Context window131K131K
Capabilities