---
title: "Cheaper alternatives to Llama 3.3 70B Instruct · Undominated.ai"
canonical: https://undominated.ai/alternatives/meta-llama__llama-3.3-70b-instruct/
description: "71 models score at least as high as Llama 3.3 70B Instruct and cost no more on a balanced workload. 33 give up nothing measurable; the other 38 name what they drop. Measured prices and independent scores."
---

# Cheaper alternatives to Llama 3.3 70B Instruct · Undominated.ai

> 71 models score at least as high as Llama 3.3 70B Instruct and cost no more on a balanced workload. 33 give up nothing measurable; the other 38 name what they drop. Measured prices and independent scores.

# Cheaper alternatives to Llama 3.3 70B Instruct

Meta · LMArena 1274.9 · $0.710/M on a balanced workload · prices as of 2026-08-28

## 33 models are both better and cheaper, giving up nothing.

Of the 133 models carrying both an independent score and a published price at the same delivery mode, **71** score at least as high as Llama 3.3 70B Instruct and cost no more under at least one workload. **33** match or beat it on every capability we hold — context, maximum output, input modes, tool use and extended reasoning. **38** do not, and every row names what it drops.

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 from that comparison and nothing else. When no model is both better and cheaper than Llama 3.3 70B Instruct, it is not generated.

## Where it holds

| Workload | Llama 3.3 70B Instruct $/M | Better and cheaper | With a trade |
| --- | --- | --- | --- |
| Balanced | $0.710 | 24 | 25 |
| Summarise | $0.710 | 33 | 38 |
| Chat | $0.710 | 23 | 22 |
| Code gen | $0.710 | 14 | 19 |
| Agentic | $0.710 | 32 | 35 |

Workload mixes are defined on the [methodology page](/methodology/). 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

| Intelligence | 1274.9 |
| --- | --- |
| Context window | 131K |
| Max output | 115K |
| Input modes | text |
| Tool use | yes |
| Extended reasoning | no |

A replacement has to clear every line above, not just the score. [Full record for Llama 3.3 70B Instruct](/models/meta-llama__llama-3.3-70b-instruct/).

## Better and cheaper, nothing given up 33

Each of these matches or beats Llama 3.3 70B Instruct on context, maximum output, input modes, tool use and reasoning, scores at least as high, and costs no more.

| # | Model | Intelligence | $/M | Holds under |
| --- | --- | --- | --- | --- |
| 1 | MiMo-V2.5-Pro Xiaomi | 1465 +190.1 | $0.544 | Balanced −23% Summarise −53% Chat −32% Code gen −7% Agentic −66% |
| 2 | Qwen3.7 Plus Qwen | 1456.2 +181.3 | $0.560 | Balanced −21% Summarise −58% Chat −12% Agentic −56% |
| 3 | Hy3 Tencent | 1441.2 +166.3 | $0.144 | Balanced −80% Summarise −89% Chat −77% Code gen −68% Agentic −88% |
| 4 | DeepSeek V4 Flash 0423 DeepSeek | 1431.6 +156.7 | $0.101 | Balanced −86% Summarise −91% Chat −87% Code gen −82% Agentic −92% |
| 5 | MiniMax M3 MiniMax | 1434.8 +159.9 | $0.525 | Balanced −26% Summarise −61% Chat −17% Agentic −59% |
| 6 | MiMo-V2.5 Xiaomi | 1427.3 +152.4 | $0.175 | Balanced −75% Summarise −85% Chat −78% Code gen −70% Agentic −89% |
| 7 | GPT-5.6 Luna OpenAI | 1428.5 +153.6 | $0.450 | Balanced −37% Summarise −72% Chat −23% Agentic −66% |
| 8 | DeepSeek V3.2 DeepSeek | 1424.6 +149.7 | $0.290 | Balanced −59% Summarise −68% Chat −62% Code gen −55% Agentic −72% |
| 9 | Qwen3 Next 80B A3B Instruct Qwen | 1418.6 +143.7 | $0.350 | Balanced −51% Summarise −80% Chat −31% Code gen −2% Agentic −67% |
| 10 | Inkling Small Thinking Machines | 1411.7 +136.8 | $0.637 | Balanced −10% Summarise −45% Chat −9% Agentic −50% |
| 11 | MiniMax M2.7 MiniMax | 1405.3 +130.4 | $0.525 | Balanced −26% Summarise −61% Chat −17% Agentic −59% |
| 12 | Qwen3.5-35B-A3B Qwen | 1395.6 +120.7 | $0.500 | Balanced −30% Summarise −58% Chat −8% Agentic −44% |
| 13 | Solar Pro 4 Upstage | 1376.2 +101.3 | $0.052 | Balanced −93% Summarise −96% Chat −92% Code gen −88% Agentic −96% |
| 14 | DeepSeek V3 0324 DeepSeek | 1375 +100.1 | $0.438 | Balanced −38% Summarise −60% Chat −23% Code gen −1% Agentic −49% |
| 15 | GPT-5.4 Nano OpenAI | 1372.8 +97.9 | $0.463 | Balanced −35% Summarise −72% Chat −20% Agentic −65% |
| 16 | gpt-oss-120b OpenAI | 1365.6 +90.7 | $0.070 | Balanced −90% Summarise −94% Chat −87% Code gen −84% Agentic −92% |
| 17 | GPT-5 Mini OpenAI | 1373.4 +98.5 | $0.688 | Balanced −3% Summarise −61% Agentic −47% |
| 18 | Gemma 3 27B Google | 1358.3 +83.4 | $0.172 | Balanced −76% Summarise −88% Chat −70% Code gen −58% Agentic −84% |
| 19 | MiniMax M2.5 MiniMax | 1359 +84.1 | $0.472 | Balanced −33% Summarise −66% Chat −27% Agentic −65% |
| 20 | Trinity Large Thinking Arcee | 1341.9 +67 | $0.378 | Balanced −47% Summarise −71% Chat −40% Code gen −18% Agentic −69% |
| 21 | MiniMax M2 MiniMax | 1342.1 +67.2 | $0.446 | Balanced −37% Summarise −59% Chat −21% Agentic −48% |
| 22 | GPT-5 Nano OpenAI | 1320.3 +45.4 | $0.138 | Balanced −81% Summarise −92% Chat −75% Code gen −64% Agentic −89% |
| 23 | Granite 4.1 8B IBM | 1291.6 +16.7 | $0.063 | Balanced −91% Summarise −93% Chat −90% Code gen −89% Agentic −92% |
| 24 | gpt-oss-20b OpenAI | 1287.8 +12.9 | $0.055 | Balanced −92% Summarise −95% Chat −90% Code gen −87% Agentic −94% |
| 25 | Kimi K2.5 Moonshot AI | 1445.2 +170.3 | $0.432 summarise | Summarise −39% Agentic −30% |
| 26 | GLM 5 Z.ai | 1445.2 +170.3 | $0.529 summarise | Summarise −25% Agentic −28% |
| 27 | Qwen3.8 27B Qwen | 1440.8 +165.9 | $0.434 summarise | Summarise −39% Agentic −24% |
| 28 | GLM 4.6 Z.ai | 1439.8 +164.9 | $0.461 summarise | Summarise −35% Agentic −31% |
| 29 | DeepSeek V4 Pro 0423 DeepSeek | 1439.2 +164.3 | $0.686 summarise | Summarise −3% Agentic −26% |
| 30 | GLM 4.7 Z.ai | 1435.3 +160.4 | $0.376 summarise | Summarise −47% Agentic −42% |
| 31 | Qwen3.5-122B-A10B Qwen | 1417.9 +143 | $0.351 summarise | Summarise −51% Agentic −25% |
| 32 | DeepSeek V3.1 DeepSeek | 1419.1 +144.2 | $0.605 summarise | Summarise −15% |
| 33 | Qwen3 235B A22B Thinking 2507 Qwen | 1413.8 +138.9 | $0.334 summarise | Summarise −53% Agentic −24% |

## Cheaper and higher-scoring, but you give something up 38

These score at least as high and cost no more on the two plotted axes, and lose something that is not on them. Read the last column before switching.

| Model | Intelligence | $/M | Holds under | What you give up |
| --- | --- | --- | --- | --- |
| Gemma 4 31B Google | 1441.7 +166.8 | $0.152 | Balanced −79% Summarise −87% Chat −75% Code gen −67% Agentic −85% | 115K → 16K max output |
| Gemma 4 26B A4B Google | 1434.6 +159.7 | $0.138 | Balanced −81% Summarise −88% Chat −75% Code gen −67% Agentic −84% | 115K → 16K max output |
| DeepSeek V3.2 Exp DeepSeek | 1424.4 +149.5 | $0.305 | Balanced −57% Summarise −61% Chat −54% Code gen −50% Agentic −59% | 115K → 66K max output |
| Qwen3 235B A22B Instruct 2507 Qwen | 1419.3 +144.4 | $0.205 | Balanced −71% Summarise −84% Chat −61% Code gen −48% Agentic −78% | 115K → 16K max output |
| DeepSeek V3.1 Terminus DeepSeek | 1419.6 +144.7 | $0.453 | Balanced −36% Summarise −62% Chat −27% Code gen −2% Agentic −58% | 115K → 33K max output |
| Qwen3 VL 235B A22B Instruct Qwen | 1420.9 +146 | $0.632 | Balanced −11% Summarise −63% Agentic −44% | 115K → 33K max output |
| Gemini 3.1 Flash Lite Preview Google | 1414.8 +139.9 | $0.563 | Balanced −21% Summarise −65% Chat −4% Agentic −57% | 115K → 66K max output |
| Step 3.5 Flash StepFun | 1403.8 +128.9 | $0.150 | Balanced −79% Summarise −85% Chat −75% Code gen −69% Agentic −82% | 115K → 66K max output |
| Qwen3.5-27B Qwen | 1407.9 +133 | $0.536 | Balanced −24% Summarise −63% Agentic −44% | 115K → 66K max output |
| Qwen3.5-Flash Qwen | 1397.6 +122.7 | $0.114 | Balanced −84% Summarise −89% Chat −80% Code gen −74% Agentic −87% | 115K → 66K max output |
| Qwen3 30B A3B Instruct 2507 Qwen | 1384.3 +109.4 | $0.084 | Balanced −88% Summarise −92% Chat −85% Code gen −81% Agentic −90% | 115K → 32K max output |
| GLM 4.5 Air Z.ai | 1382.8 +107.9 | $0.310 | Balanced −56% Summarise −81% Chat −46% Code gen −22% Agentic −75% | 115K → 98K max output |
| GLM 4.6V Z.ai | 1374.7 +99.8 | $0.450 | Balanced −37% Summarise −63% Chat −34% Code gen −10% Agentic −66% | 115K → 33K max output |
| Qwen3 Next 80B A3B Thinking Qwen | 1367.5 +92.6 | $0.412 | Balanced −42% Summarise −71% Chat −20% Agentic −57% | 115K → 33K max output |
| Mercury 2 Inception | 1357.8 +82.9 | $0.375 | Balanced −47% Summarise −70% Chat −46% Code gen −25% Agentic −73% | 131K → 128K context 115K → 50K max output |
| GLM 4.7 Flash Z.ai | 1352.9 +78 | $0.145 | Balanced −80% Summarise −91% Chat −75% Code gen −63% Agentic −89% | 115K → 16K max output |
| Qwen3 32B Qwen | 1340.1 +65.2 | $0.130 | Balanced −82% Summarise −87% Chat −77% Code gen −72% Agentic −85% | 115K → 16K max output |
| Gemma 3 12B Google | 1334.2 +59.3 | $0.075 | Balanced −89% Summarise −92% Chat −87% Code gen −85% Agentic −91% | 115K → 16K max output |
| GPT-4.1 Mini OpenAI | 1340.5 +65.6 | $0.700 | Balanced −1% Summarise −47% Agentic −43% | 115K → 33K max output |
| DeepSeek V3 DeepSeek | 1332.6 +57.7 | $0.450 | Balanced −37% Summarise −58% Chat −20% Agentic −47% | 115K → 16K max output |
| Qwen3 30B A3B Qwen | 1316.9 +42 | $0.215 | Balanced −70% Summarise −80% Chat −62% Code gen −51% Agentic −75% | 115K → 16K max output |
| Olmo 3 32B Think Allen AI | 1298.5 +23.6 | $0.237 | Balanced −67% Summarise −76% Chat −59% Code gen −49% Agentic −71% | 131K → 66K context 115K → 59K max output no tool use |
| Gemma 3 4B Google | 1290.8 +15.9 | $0.063 | Balanced −91% Summarise −93% Chat −90% Code gen −89% Agentic −92% | 115K → 16K max output no tool use |
| GPT-4o-mini (2024-07-18) OpenAI | 1286.6 +11.7 | $0.263 | Balanced −63% Summarise −79% Chat −57% Code gen −42% Agentic −76% | 131K → 128K context 115K → 16K max output |
| GPT-4.1 Nano OpenAI | 1284.8 tie | $0.175 | Balanced −75% Summarise −87% Chat −72% Code gen −61% Agentic −86% | 115K → 33K max output |
| Gemini 3.7 Flash Google | 1490.2 +215.3 | $0.354 summarise | Summarise −50% Agentic −44% | 115K → 66K max output |
| Gemini 3.6 Flash Google | 1476.5 +201.6 | $0.708 summarise | Summarise under 1% less | 115K → 66K max output |
| Gemini 3.5 Flash Lite Google | 1436.5 +161.6 | $0.333 summarise | Summarise −53% Agentic −34% | 115K → 66K max output |
| Qwen3.5 397B A17B Qwen | 1438.3 +163.4 | $0.487 summarise | Summarise −31% Agentic −4% | 115K → 66K max output |
| Qwen3.6 Plus Qwen | 1436.8 +161.9 | $0.406 summarise | Summarise −43% Agentic −20% | 115K → 66K max output |
| GLM 4.5 Z.ai | 1429.4 +154.5 | $0.540 summarise | Summarise −24% Agentic −23% | 115K → 98K max output |
| R1 0528 DeepSeek | 1427.9 +153 | $0.540 summarise | Summarise −24% Agentic −7% | 115K → 33K max output |
| Gemini 2.5 Flash Google | 1417.3 +142.4 | $0.333 summarise | Summarise −53% Agentic −34% | 115K → 66K max output |
| Qwen3 VL 235B A22B Thinking Qwen | 1400.6 +125.7 | $0.580 summarise | Summarise −18% | 115K → 33K max output |
| Qwen3 235B A22B Qwen | 1366 +91.1 | $0.523 summarise | Summarise −26% Agentic −7% | 115K → 8K max output |
| Nova 2 Lite Amazon | 1362.2 +87.3 | $0.410 summarise | Summarise −42% Agentic −11% | 115K → 66K max output |
| MiniMax M1 MiniMax | 1341.9 +67 | $0.632 summarise | Summarise −11% | 115K → 40K max output |
| GLM 4.5V Z.ai | 1333.6 +58.7 | $0.520 summarise | Summarise −27% Agentic −31% | 131K → 66K context 115K → 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 ±10.68 points, so a gap smaller than that is marked *tie* rather than an improvement — see [significance bands](/significance/).
 - 188 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.
 - The models that nothing beats on both axes are on the [value frontier](/frontier/), and every other model something cheaper beats is [listed here](/alternatives/).