DeepSeek-R1-Distill-Llama-8B vs Granite 3.3 8B Instruct
Wins 1 of 4 areas
Coding
Wins 3 of 4 areas
Reasoning · Long documents · Following instructions
Granite 3.3 8B Instruct is the stronger all-rounder.DeepSeek-R1-Distill-Llama-8B is cheaper and better at coding.
Scores updated · 18 tests both models report · How we compare
Where each one wins
Tests won in each of the four areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking01Granite 3.3 8B Instruct1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts01Granite 3.3 8B Instruct1 of 1 test
- Following instructionsDoing exactly what it is asked01Granite 3.3 8B Instruct1 of 1 test
- CodingWriting and fixing software10DeepSeek-R1-Distill-Llama-8B1 of 1 test
Coding, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
DeepSeek-R1-Distill-Llama-8B costs 64% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Granite 3.3 8B Instruct pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+4.8points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+3.6points ahead
Where DeepSeek-R1-Distill-Llama-8B pulls ahead
- Code for real scientific research problemsSciCode+1.8points ahead
Every test, side by side
All 18 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-R1-Distill-Llama-8B
- SciCodeDeepSeek-R1-Distill-Llama-8B by 1.811.910.1+1.8
ReasoningGranite 3.3 8B Instruct
- GPQA DiamondGranite 3.3 8B Instruct by 3.630.233.8+3.6
- Humanity's Last Examtie4.24.2tie
Long documentsGranite 3.3 8B Instruct
- AA-LCRGranite 3.3 8B Instruct by 3.303.3+3.3
Following instructionsGranite 3.3 8B Instruct
- IFBenchGranite 3.3 8B Instruct by 4.817.622.4+4.8
Other results13 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AttaQGranite 3.3 8B Instruct by 45.642.988.5+45.6
- AIME 2024DeepSeek-R1-Distill-Llama-8B by 42.350.48.1+42.3
- AlpacaEval-2.0Granite 3.3 8B Instruct by 40.821.962.7+40.8
- Arena HardGranite 3.3 8B Instruct by 40.417.257.6+40.4
- HumanEval+Granite 3.3 8B Instruct by 23.262.986.1+23.2
- HumanEvalGranite 3.3 8B Instruct by 22.267.589.7+22.2
- MATH-500 (EM)DeepSeek-R1-Distill-Llama-8B by 20.189.169+20.1
- TruthfulQAGranite 3.3 8B Instruct by 19.547.466.9+19.5
- PopQAGranite 3.3 8B Instruct by 12.913.326.2+12.9
- DROPGranite 3.3 8B Instruct by 9.749.759.4+9.7
- GSM8KGranite 3.3 8B Instruct by 8.772.280.9+8.7
- BigBenchHardGranite 3.3 8B Instruct by 1.767.469.1+1.7
- AA IntelligenceDeepSeek-R1-Distill-Llama-8B by 1.66.54.9+1.6
Questions people ask
Which is better, DeepSeek-R1-Distill-Llama-8B or Granite 3.3 8B Instruct?
Granite 3.3 8B Instruct wins three of the four areas where both have results: reasoning, long documents and following instructions. DeepSeek-R1-Distill-Llama-8B wins coding, and costs 64% less.
Which is better for coding?
DeepSeek-R1-Distill-Llama-8B. It wins the one coding test both models report.
Which is cheaper?
DeepSeek-R1-Distill-Llama-8B costs $0.05 per million input tokens and $0.05 per million output tokens; Granite 3.3 8B Instruct costs $0.03 and $0.25. That makes DeepSeek-R1-Distill-Llama-8B about 64% cheaper for the same work.
How do you compare the two?
We use the 18 benchmark tests both models have published scores on. The verdict counts the 5 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 13 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.