1stTop pick
Researched 29d agoGemma 2 9B
- MMLU-Pro
- 52.08%
- Output (from)
- $0.180 / 1M
Best Small Language Models Under 10B Parameters (2026)
Last refreshed 2026-05-16. Next refresh: weekly.
Efficient small language models for edge deployment, cost-sensitive workloads, or on-device inference. Under 10B parameters with strong benchmark scores.
Top three picks
Opinionated short stack for this category — scroll for the full leaderboard, pricing, and compare links.
2ndShortlist
Researched 29d agoLlama 3.1 8B Instruct
- MMLU-Pro
- 44.25%
- Output (from)
- $0.050 / 1M
3rdShortlist
Researched 29d agoLlama 3 8B Instruct
- MMLU-Pro
- 40.5%
- Output (from)
- $0.040 / 1M
How we rank
Small models (≤10B active parameters) rank on MMLU-Pro, then GPQA Diamond, MMLU, and HellaSwag.
- Eligibility — Non-deprecated models with ≤10B parameters (billions-only parser).
- Primary ranking — MMLU-Pro, then GPQA Diamond, then MMLU, then HellaSwag, then newer release.
- Podium freshness — Shortlist cards require `lastResearched` within 60 days and a tracked public output price. Stale or unpriced SKUs stay in the table with a “Verify pricing” badge once research is past 45 days.
- Variant collapse — We keep one row per model family (`familySlug` + parameter tier). When headline scores tie within ±0.5 pt (±10 Elo on Chatbot Arena), we pick the canonical SKU by lowest tracked input price, then GA over preview or limited access, then newest `release`. A folded sibling within the benchmark noise band can show a "Tied within margin" chip on that score cell.
- Pricing — SLMs often win on unit economics — compare the provider ladder before picking.
| # | Model | Input $/1M | Output $/1M | |
|---|---|---|---|---|
| 1 | Phi-4 Mini MMLU-Pro: 52.8% | $0.05 | $0.15 | |
| 2 | Gemma 2 9B MMLU-Pro: 52.08% | $0.06 | $0.18 | |
| 3 | Phi-3 Mini 4k MMLU-Pro: 45.66% | $0.05 | $0.25 | |
| 4 | Llama 3.1 8B Instruct MMLU-Pro: 44.25% | $0.02 | $0.05 | |
| 5 | Llama 3 8B Instruct MMLU-Pro: 40.5% | $0.03 | $0.04 | |
| 6 | Llama 3.2 3B Instruct MMLU-Pro: 34.7% | $0.05 | $0.10 | |
| 7 | Llama 3.2 1B Instruct MMLU-Pro: 20% | $0.03 | $0.10 | |
| 8 | MiniMax M2.7 ReasoningTools MMLU-Pro: — | $0.30 | $1.20 | |
| 9 | Qwen3-8B MMLU-Pro: — | $0.05 | $0.20 | |
| 10 | Qwen2-7B MMLU-Pro: — | $0.05 | $0.15 | |
| 11 | Gemma 7B Instruct MMLU-Pro: — | $0.05 | $0.07 | |
| 12 | OpenChat 3.5 (0106) MMLU-Pro: — | $0.07 | $0.07 | |
| 13 | Starling LM 7B Beta MMLU-Pro: — | — | — | |
| 14 | Zephyr 7B Beta MMLU-Pro: — | $0.05 | $0.20 | |
| 15 | Qwen2.5-7B-Instruct MMLU-Pro: — | $0.03 | $0.03 | |
| 16 | Aya 23 8B MMLU-Pro: — | — | — | |
| 17 | Qwen2-1.5B MMLU-Pro: — | $0.07 | $0.07 | |
| 18 | Qwen1.5-7B MMLU-Pro: — | $0.05 | $0.20 | |
| 19 | Phi-2 MMLU-Pro: — | $0.05 | $0.07 | |
| 20 | Gemma 2B MMLU-Pro: — | — | — |
Honorable mentions
Next seats in this ranking. Lines below are from each model's stored description in LLMReference seed data—spot-check the model page before relying on a capability claim.
Llama 3.2 3B Instruct available on AWS Bedrock
34.7%
MMLU-Pro
Llama 3.2 1B Instruct available on AWS Bedrock
20%
MMLU-Pro
MiniMax M2.7 is MiniMax's self-improving frontier model, released March 18, 2026. It introduces native multi-agent collaboration, complex skill orchestration, and early recursive self-improvement capabilities. The model uses 10B active parameters, supports a 204,800-token context window, and was released alongside MiniMax-M2.7-highspeed, a 66% faster latency-optimized variant. Public provider listings price standard M2.7 at $0.30 per 1M input tokens and $1.20 per 1M output tokens.
—
MMLU-Pro
Compare Top Picks
Browse Other Categories
Best LLMs for Code GenerationBest LLMs for RAGBest AI Agents & Agentic ModelsBest LLMs for ClassificationBest Open Source LLMsBest Multimodal / Vision LLMsBest LLMs for Reasoning & MathBest LLMs for Function Calling & Tool UseCheapest LLM APIs You Can Call Right NowBest Long Context LLMsBest Mainstream LLM APIs, RankedBest LLMs for EnterpriseBest Free LLMs You Can Use Right NowBest LLMs for WritingBest LLMs for MarketingBest LLMs for Customer Support