Compare/Devstral Small (May '25) vs LFM2 24B A2B

Devstral Small (May '25)vsLFM2 24B A2B

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

Mistral

Devstral Small (May '25)

Input
$0.06/M
Output
$0.12/M
Speed
TTFT
Liquid AI

LFM2 24B A2B

Input
$0.03/M
Output
$0.12/M
Speed
211 tok/s
TTFT
0.22s

Winner by Category

Cheaper
LFM2 24B A2B
Faster (tok/s)
LFM2 24B A2B
Lower Latency
Devstral Small (May '25)
Benchmarks (7-3)
Devstral Small (May '25)

Pricing Comparison

MetricDevstral Small (May '25)LFM2 24B A2B
Input ($/M tokens)$0.06$0.03
Output ($/M tokens)$0.12$0.12
Cost for 1M input + 100K output tokens:
Devstral Small (May '25)$0.07
LFM2 24B A2B$0.04

Speed Comparison

Output Speed (tokens/s) — higher is better
Devstral Small (May '25)
LFM2 24B A2B
211 tok/s
Time to First Token (seconds) — lower is better
Devstral Small (May '25)
LFM2 24B A2B
0.22s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
18.010.5
Coding Index
12.23.6
Math Index
GPQA Diamond
43.4%47.4%
MMLU-Pro
63.2%
LiveCodeBench
25.8%
AIME 2025
MATH-500
68.4%
Humanity's Last Exam
4.0%4.4%
SciCode
24.5%10.9%
IFBench
31.6%45.9%
TerminalBench
6.1%0.0%
Devstral Small (May '25)7 wins
3 winsLFM2 24B A2B

Frequently Asked Questions

Which is cheaper, Devstral Small (May '25) or LFM2 24B A2B?

LFM2 24B A2B is cheaper overall. Its blended price (3:1 input/output ratio) is $0.05/M tokens vs $0.07/M for Devstral Small (May '25).

Which model performs better on benchmarks?

Devstral Small (May '25) wins 7 out of 12 benchmarks compared to 3 for LFM2 24B A2B. See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

LFM2 24B A2B generates tokens faster at 211 tok/s vs 0 tok/s. Devstral Small (May '25) also has lower time-to-first-token (0.00s vs 0.22s).

When should I use Devstral Small (May '25) vs LFM2 24B A2B?

Choose based on your priorities: LFM2 24B A2B for lower cost, Devstral Small (May '25) for stronger benchmark performance, and LFM2 24B A2B for faster generation. For latency-sensitive apps, check the TTFT comparison above.