IBM Granite 4.2 3B vs IBM Granite 4.2 8B: Specs & Benchmark Comparison

IBM Granite 4.2 3B and IBM Granite 4.2 8B both come from IBM, and both were released in August 2026. IBM Granite 4.2 8B is the larger model at roughly 8 billion parameters, against 3 billion for IBM Granite 4.2 3B.

The two models share 13 published benchmarks. IBM Granite 4.2 8B leads on 12 of them, IBM Granite 4.2 3B on 1. The widest gaps are on Arena-Hard v2, where IBM Granite 4.2 8B scores 65.2% against 35.0%; RULER 128k, where IBM Granite 4.2 8B scores 71.4% against 55.3%. Averaged across everything we track, IBM Granite 4.2 3B sits at 55.7% and IBM Granite 4.2 8B at 56.7%.

IBM Granite 4.2 3B is available through an API at $0.03 per million input tokens with a 131K-token context window. We do not track a hosted API for IBM Granite 4.2 8B, so pricing and context cannot be compared directly.

CharacteristicIBM Granite 4.2 3BIBM Granite 4.2 8B
CompanyIBMIBM
Release DateAugust 25, 2026August 25, 2026
Parameters3B8B
MultimodalNoNo
Context (input)131K
Context (output)131K
Input Price / 1M$0.03
Output Price / 1M$0.12
Average Score55.7%56.7%
Benchmarks
Arena-Hard v235.0%65.2%
RULER 128k55.3%71.4%
RULER 64k67.5%81.0%
TAU3-Bench45.8%58.1%
SciCode24.1%36.1%
HMMT2566.7%78.3%
GPQA54.8%64.1%
ProfBench32.1%41.2%
AIME 202578.3%86.7%
MMLU-Pro67.8%74.0%
IFBench74.3%79.3%
LiveCodeBench v669.7%73.2%
BFCL-V452.4%52.4%

Visual Benchmark Comparison

IBM Granite 4.2 3B
IBM Granite 4.2 8B
Arena-Hard v20.3 vs 0.7
0.3
0.7
RULER 128k0.6 vs 0.7
0.6
0.7
RULER 64k0.7 vs 0.8
0.7
0.8
TAU3-Bench0.5 vs 0.6
0.5
0.6
SciCode0.2 vs 0.4
0.2
0.4
HMMT250.7 vs 0.8
0.7
0.8
GPQA0.5 vs 0.6
0.5
0.6
ProfBench0.3 vs 0.4
0.3
0.4
AIME 20250.8 vs 0.9
0.8
0.9
MMLU-Pro0.7 vs 0.7
0.7
0.7
IFBench0.7 vs 0.8
0.7
0.8
LiveCodeBench v60.7 vs 0.7
0.7
0.7
BFCL-V40.5 vs 0.5
0.5
0.5

Verdict

Both models show equal results — the choice depends on your specific use case.

Overall Performance

Both models show comparable average scores: IBM Granite 4.2 3B — 0.6, IBM Granite 4.2 8B — 0.6.

Recency

Both models were released around the same time: 8/25/2026 and 8/25/2026.

More About These Models

Related Comparisons

Frequently Asked Questions

Which is better for coding — IBM Granite 4.2 3B or IBM Granite 4.2 8B?
Direct comparison on the SWE-Bench benchmark is not available. We recommend reviewing other metrics on the comparison page.
Which model is cheaper — IBM Granite 4.2 3B or IBM Granite 4.2 8B?
API pricing data is available on the individual model pages.
Which has a larger context window — IBM Granite 4.2 3B or IBM Granite 4.2 8B?
Context window data is available on the individual model pages.

The IBM Granite 4.2 3B and IBM Granite 4.2 8B comparison is updated for 2026. Data includes benchmark results, API pricing, context window size and other specifications. For more detailed information, visit the IBM Granite 4.2 3B or IBM Granite 4.2 8B page. See also the complete list of AI model comparisons.