1stTop pick
Researched 19d agoQwen3.5-9B
- MMLU-Pro
- 82.5%
- Output (from)
- $0.150 / 1M
Best Small Language Models Under 10B Parameters (2026)
As of 2026-11-13, LLM Reference's top pick for small-model deployments is Qwen3.5-9B by Alibaba (MMLU-Pro 82.5%, $0.150/1M output tokens), chosen from models that meet the podium eligibility rules. Second pick: Granite 4.2 8B by IBM Research.
Last refreshed 2026-11-13. Next refresh: weekly.
The best small LLMs under 10B parameters in 2026 — fast, cheap, and deployable on-device or at the edge with strong benchmark scores.
Verdict
Use Qwen3.5-9B for small-model deployments today.
Granite 4.2 8B is the runner-up, 8 points back on MMLU-Pro.
2ndShortlist
Researched 40d agoGranite 4.2 8B
- MMLU-Pro
- 74.04%
- Output (from)
- $0.250 / 1M
3rdShortlist
Researched 30d agoClef-flash
- MMLU-Pro
- 65.3%
- Output (from)
- Free
How we rank
Small models (at most 10B total parameters, not active parameters) rank on MMLU-Pro, then GPQA Diamond, MMLU, and HellaSwag.
- Eligibility — Non-deprecated models with at most 10B total parameters.
- Primary ranking — MMLU-Pro, then GPQA Diamond, then MMLU, then HellaSwag, then newer release.
- Podium freshness — Shortlist cards require a model researched within the last 60 days and a listed public output price. Older or unpriced SKUs stay in the table with a “Verify pricing” badge when research is older than 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.
- Podium eligibility — Preview, restricted, retired and retiring models never fill a podium slot. Preview means preview or invite-only access, or a model that is announced or pre-release. Restricted means restricted, partner, private or otherwise unavailable access. Retired means the model is past its retirement date or every provider route is retired. Retiring means the model, or any one of its provider routes, has a retirement date within the next 60 days. These models can still appear in the table below the podium.
- Pricing — SLMs often win on unit economics — compare the provider ladder before picking.
| # | Model | Input $/1M | Output $/1M | |
|---|---|---|---|---|
| 1 | Qwen3.5-9B VisionTools MMLU-Pro: 82.5% | $0.10 | $0.15 | |
| 2 | Qwen3.5-4B VisionVerify pricing MMLU-Pro: 79.1% | — | — | |
| 3 | ZAYA1-8B ReasoningVerify pricing MMLU-Pro: 74.2% | — | — | |
| 4 | Granite 4.2 8B ReasoningTools MMLU-Pro: 74.04% | $0.06 | $0.25 | |
| 5 | Gemma 4 E4B IT ToolsVerify pricing MMLU-Pro: 69.4% | Free | Free | |
| 6 | Granite 4.2 3B ReasoningTools MMLU-Pro: 67.84% | — | — | |
| 7 | Qwen3.5-2B VisionVerify pricing MMLU-Pro: 66.5% | — | — | |
| 8 | Clef-flash Vision MMLU-Pro: 65.3% | $0.04 | Free | |
| 9 | Gemma 4 E2B IT ToolsVerify pricing MMLU-Pro: 60% | Free | Free | |
| 10 | Phi-4 Mini MMLU-Pro: 52.8% | $0.07 | $0.30 | |
| 11 | Gemma 2 9B Verify pricing MMLU-Pro: 52.08% | $0.06 | $0.18 | |
| 12 | LFM2.5 8B A1B ReasoningToolsVerify pricing MMLU-Pro: 50.5% | — | — | |
| 13 | Granite 4.1 3B Verify pricing MMLU-Pro: 49.83% | — | — | |
| 14 | Phi-3 Mini 4k Verify pricing MMLU-Pro: 45.66% | $0.05 | $0.25 | |
| 15 | Granite 4.0 H Tiny ToolsVerify pricing MMLU-Pro: 44.94% | — | — | |
| 16 | LFM2.5 1.2B Instruct ToolsVerify pricing MMLU-Pro: 44.35% | — | — | |
| 17 | Llama 3.1 8B Instruct MMLU-Pro: 44.25% | $0.02 | $0.05 | |
| 18 | Llama 3 8B Instruct MMLU-Pro: 40.5% | $0.05 | $0.20 | |
| 19 | Llama 3.2 3B Instruct Verify pricing MMLU-Pro: 34.7% | $0.03 | $0.05 | |
| 20 | Llama 3.2 1B Instruct Verify pricing MMLU-Pro: 20% | $0.03 | $0.10 |
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.
Instruction-tuned variant of Gemma 4 E2B for edge deployment with enhanced instruction-following capabilities.
60%
MMLU-Pro
Phi-4 family model from Microsoft Research. Mini variant with efficient performance.
52.8%
MMLU-Pro
Gemma 2 9B is Google DeepMind's Gemma 2 model. It offers an 8K-token context window with weights openly available for self-hosting and scores 52.1 on MMLU PRO.
52.08%
MMLU-Pro
Compare Top Picks
Side-by-side comparison of the top picks by price, benchmark, and API access.
Browse Other Categories
Best LLMs for Code GenerationBest LLMs for RAGBest LLMs for Agents vs Agent Products (2026)Best LLMs for ClassificationBest Open Source LLMsBest Multimodal / Vision LLMsBest LLM for Translation in 2026Best AI Image Models in 2026Best AI Video Models in 2026Best LLMs for Reasoning & MathBest LLMs for JSON / 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
Frequently asked questions
What is the best LLM for small-model deployments?
LLM Reference's top pick for small-model deployments as of 2026-11-13 is Qwen3.5-9B by Alibaba, with MMLU-Pro 82.5%. Picks come from models that meet the podium eligibility rules, so the top pick may not lead the full table. Second pick: Granite 4.2 8B by IBM Research (MMLU-Pro 74.04%).
Which ranked LLM for small-model deployments has the lowest output price?
Among the ranked models shown on this page, Gemma 4 E4B IT by Google DeepMind has the lowest listed output price: free, at rank 5.
What is the best open-weight LLM for small-model deployments?
Qwen3.5-9B by Alibaba is the highest-ranked open-weight model shown on this page (rank 1, MMLU-Pro 82.5%). License: Apache 2.0.
How is the small-model deployments ranking calculated?
Small models (at most 10B total parameters, not active parameters) rank on MMLU-Pro, then GPQA Diamond, MMLU, and HellaSwag. Preview, restricted, retired and retiring models never fill a podium slot. Preview means preview or invite-only access, or a model that is announced or pre-release. Restricted means restricted, partner, private or otherwise unavailable access. Retired means the model is past its retirement date or every provider route is retired. Retiring means the model, or any one of its provider routes, has a retirement date within the next 60 days. These models can still appear in the table below the podium.
How often is this page updated?
Weekly. Last refreshed 2026-11-13.