Undominated.ai

Alternatives to Llama 3.2 3B Instruct

Meta LMArena 1109.5 $0.120/M on a balanced workload prices as of 2026-10-06

11 models are higher-scoring or cheaper, with neither dimension worse.

Of the 144 models carrying both an independent score and a published price at the same delivery mode, 18 score at least as high as Llama 3.2 3B Instruct and cost no more under at least one listed workload, with at least one of those dimensions improved. 11 match or beat its recorded context, maximum output, input modes, tool use and extended reasoning. 7 have a recorded capability loss, named on every row.

Dominance is not a property of a model. It is a property of a model, a capability lens and a workload, and all three are named on every row. Scores are LMArena Elo, used under CC BY 4.0 from the official dataset; prices are blended per million tokens. A higher score is not a drop-in replacement.

This page is built only while a candidate improves score or price without worsening the other under a listed workload. When that comparison no longer holds, it is not generated.

Where it holds

WorkloadLlama 3.2 3B Instruct $/MRecorded capabilities preservedRecorded capability losses
Balanced$0.12056
Summarise$0.06454
Chat$0.16287
Code gen$0.21897
Agentic$0.092105

Workload mixes are defined on the methodology page. Cached input is priced at the cached rate, and reasoning tokens at the worse of the reasoning and output rates.

What you would be replacing

Intelligence1109.5
Context window131K
Max output117K
Input modestext
Tool useno
Extended reasoningno

Preserving recorded capabilities requires matching or beating these fields. Full record for Llama 3.2 3B Instruct, or check Llama 3.2 3B Instruct against the whole catalogue.

Recorded capabilities preserved 11

Each option has a higher score or a lower price, with neither dimension worse under the listed workloads. It matches or beats Llama 3.2 3B Instruct on recorded context, maximum output, input modes, tool use and extended reasoning.

#ModelIntelligence$/MHolds under
1DeepSeek V4 Flash 0423 DeepSeek1432.1 +322.6$0.113
  • Balanced −6%
  • Chat −36%
  • Code gen −37%
  • Agentic −34%
2gpt-oss-120b OpenAI1365.4 +255.9$0.065
  • Balanced −46%
  • Summarise −42%
  • Chat −47%
  • Code gen −48%
  • Agentic −45%
3Gemma 3 27B Google1357.8 +248.3$0.100
  • Balanced −17%
  • Chat −31%
  • Code gen −41%
  • Agentic same price
4Granite 4.2 8B IBM1317.8 +208.3$0.108
  • Balanced −10%
  • Summarise −11%
  • Chat −24%
  • Code gen −22%
  • Agentic −33%
5Llama 3.1 8B Instruct Meta1186.5 +77$0.025
  • Balanced −79%
  • Summarise −67%
  • Chat −83%
  • Code gen −85%
  • Agentic −75%
6MiMo-V2.6-Flash Xiaomi1456.4 +346.9$0.155 chat
  • Chat −4%
  • Code gen −2%
  • Agentic −14%
7Gemma 4 26B A4B Google1433.9 +324.4$0.149 chat
  • Chat −8%
  • Code gen −6%
  • Agentic −13%
8MiMo-V2.5 Xiaomi1427.5 +318$0.155 chat
  • Chat −4%
  • Code gen −2%
  • Agentic −14%
9Qwen3 30B A3B Instruct 2507 Qwen1384.1 +274.6$0.216 code gen
  • Code gen −1%
10GLM 4.7 Flash Z.ai1350.9 +241.4$0.063 summarise
  • Summarise −2%
  • Agentic −12%
11GPT-5 Nano OpenAI1320 +210.5$0.055 summarise
  • Summarise −15%
  • Agentic −18%

Alternatives with recorded capability losses 7

These improve score or price without worsening the other under the listed workloads, but reduce at least one recorded capability. Read the last column before switching.

ModelIntelligence$/MHolds underRecorded capability loss
Qwen3.5-Flash Qwen1397 +287.5$0.114
  • Balanced −5%
  • Chat −12%
  • Code gen −17%
  • 118K → 66K max output
Gemma 3 12B Google1334.2 +224.7$0.075
  • Balanced −38%
  • Summarise −14%
  • Chat −44%
  • Code gen −50%
  • Agentic −29%
  • 118K → 16K max output
Gemma 3 4B Google1290.8 +181.3$0.063
  • Balanced −48%
  • Summarise −18%
  • Chat −57%
  • Code gen −63%
  • Agentic −38%
  • 118K → 16K max output
gpt-oss-20b OpenAI1287.7 +178.2$0.036
  • Balanced −70%
  • Summarise −70%
  • Chat −73%
  • Code gen −72%
  • Agentic −75%
  • 118K → 33K max output
Mistral Small 3 Mistral1233.6 +124.1$0.058
  • Balanced −52%
  • Summarise −20%
  • Chat −62%
  • Code gen −69%
  • Agentic −41%
  • 131K → 33K context
  • 118K → 16K max output
Phi 4 Microsoft1216.7 +107.2$0.088
  • Balanced −27%
  • Chat −40%
  • Code gen −49%
  • Agentic −13%
  • 131K → 16K context
  • 118K → 15K max output
Qwen3 32B Qwen1340.1 +230.6$0.160 chat
  • Chat −1%
  • Code gen −8%
  • 118K → 16K max output

What this compares, and what it leaves out

  • Quality is LMArena Elo, used under CC BY 4.0 from the official dataset. The 95% confidence interval on a difference between two scores is about ±10.47 points, so a gap smaller than that is marked tie rather than an improvement — see significance bands.
  • 204 further models at this delivery mode carry a price but no independent score. They are absent from the comparison in both directions — unrated is not a zero, and an unmeasured model is neither an alternative nor a worse buy.
  • Retired models are never offered as an alternative, and a model only competes against its own delivery mode: batch trades latency for price, so it is not a like-for-like swap.
  • Nothing here measures latency, throughput, rate limits or how a model behaves on your prompts. Two models with the same index score are not interchangeable.
  • Models with no higher-scoring or cheaper alternative that is no worse on the other axis are on the value frontier. Models with such an alternative are listed here.
Evidence & Ask
Markdown for LLMs