Compare/Granite 3.3 8B (Non-reasoning) vs DeepSeek R1 Distill Qwen 32B

Granite 3.3 8B (Non-reasoning)vsDeepSeek R1 Distill Qwen 32B

Side-by-side comparison of pricing, 12 benchmarks, and generation speed.

IBM

Granite 3.3 8B (Non-reasoning)

Input
$0.03/M
Output
$0.25/M
Speed
386 tok/s
TTFT
7.20s
DeepSeek

DeepSeek R1 Distill Qwen 32B

Input
$0.27/M
Output
$0.27/M
Speed
59 tok/s
TTFT
0.49s

Winner by Category

Cheaper
Granite 3.3 8B (Non-reasoning)
Faster (tok/s)
Granite 3.3 8B (Non-reasoning)
Lower Latency
DeepSeek R1 Distill Qwen 32B
Benchmarks (2-10)
DeepSeek R1 Distill Qwen 32B

Pricing Comparison

MetricGranite 3.3 8B (Non-reasoning)DeepSeek R1 Distill Qwen 32B
Input ($/M tokens)$0.03$0.27
Output ($/M tokens)$0.25$0.27
Cost for 1M input + 100K output tokens:
Granite 3.3 8B (Non-reasoning)$0.06
DeepSeek R1 Distill Qwen 32B$0.30

Speed Comparison

Output Speed (tokens/s) — higher is better
Granite 3.3 8B (Non-reasoning)
386 tok/s
DeepSeek R1 Distill Qwen 32B
59 tok/s
Time to First Token (seconds) — lower is better
Granite 3.3 8B (Non-reasoning)
7.20s
DeepSeek R1 Distill Qwen 32B
0.49s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
7.017.2
Coding Index
3.4
Math Index
6.763.0
GPQA Diamond
33.8%61.5%
MMLU-Pro
46.8%73.9%
LiveCodeBench
12.7%27.0%
AIME 2025
6.7%63.0%
MATH-500
66.5%94.1%
Humanity's Last Exam
4.2%5.5%
SciCode
10.1%37.6%
IFBench
22.4%22.9%
TerminalBench
0.0%
Granite 3.3 8B (Non-reasoning)2 wins
10 winsDeepSeek R1 Distill Qwen 32B

Frequently Asked Questions

Which is cheaper, Granite 3.3 8B (Non-reasoning) or DeepSeek R1 Distill Qwen 32B?

Granite 3.3 8B (Non-reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.09/M tokens vs $0.27/M for DeepSeek R1 Distill Qwen 32B.

Which model performs better on benchmarks?

DeepSeek R1 Distill Qwen 32B wins 10 out of 12 benchmarks compared to 2 for Granite 3.3 8B (Non-reasoning). See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

Granite 3.3 8B (Non-reasoning) generates tokens faster at 386 tok/s vs 59 tok/s. However, DeepSeek R1 Distill Qwen 32B has lower time-to-first-token (0.49s vs 7.20s).

When should I use Granite 3.3 8B (Non-reasoning) vs DeepSeek R1 Distill Qwen 32B?

Choose based on your priorities: Granite 3.3 8B (Non-reasoning) for lower cost, DeepSeek R1 Distill Qwen 32B for stronger benchmark performance, and Granite 3.3 8B (Non-reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.