---
title: "Cheaper alternatives to Trinity Large Thinking · Undominated.ai"
canonical: https://undominated.ai/alternatives/arcee-ai__trinity-large-thinking/
description: "21 models score at least as high as Trinity Large Thinking and cost no more on a balanced workload. 1 give up nothing measurable; the other 20 name what they drop. Measured prices and independent scores."
---

# Cheaper alternatives to Trinity Large Thinking · Undominated.ai

> 21 models score at least as high as Trinity Large Thinking and cost no more on a balanced workload. 1 give up nothing measurable; the other 20 name what they drop. Measured prices and independent scores.

# Cheaper alternatives to Trinity Large Thinking

Arcee · LMArena 1341.9 · $0.378/M on a balanced workload · prices as of 2026-08-28

## 1 model is 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, **21** score at least as high as Trinity Large Thinking and cost no more under at least one workload. **1** matches or beats it on every capability we hold — context, maximum output, input modes, tool use and extended reasoning. **20** 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 Trinity Large Thinking, it is not generated.

## Where it holds

| Workload | Trinity Large Thinking $/M | Better and cheaper | With a trade |
| --- | --- | --- | --- |
| Balanced | $0.378 | 1 | 17 |
| Summarise | $0.206 | 1 | 17 |
| Chat | $0.424 | 1 | 16 |
| Code gen | $0.585 | 1 | 16 |
| Agentic | $0.219 | 1 | 15 |

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 | 1341.9 |
| --- | --- |
| Context window | 262K |
| Max output | 236K |
| Input modes | text |
| Tool use | yes |
| Extended reasoning | yes |

A replacement has to clear every line above, not just the score. [Full record for Trinity Large Thinking](/models/arcee-ai__trinity-large-thinking/).

## Better and cheaper, nothing given up 1

Each of these matches or beats Trinity Large Thinking 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 | DeepSeek V4 Flash 0423 DeepSeek | 1431.6 +89.7 | $0.101 | Balanced −73% Summarise −68% Chat −78% Code gen −79% Agentic −75% |

## Cheaper and higher-scoring, but you give something up 20

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 +99.8 | $0.152 | Balanced −60% Summarise −56% Chat −58% Code gen −60% Agentic −53% | 236K → 16K max output |
| Hy3 Tencent | 1441.2 +99.3 | $0.144 | Balanced −62% Summarise −63% Chat −62% Code gen −61% Agentic −62% | 236K → 128K max output |
| Gemma 4 26B A4B Google | 1434.6 +92.7 | $0.138 | Balanced −64% Summarise −59% Chat −58% Code gen −60% Agentic −50% | 236K → 16K max output |
| MiMo-V2.5 Xiaomi | 1427.3 +85.4 | $0.175 | Balanced −54% Summarise −48% Chat −63% Code gen −64% Agentic −64% | 236K → 131K max output |
| DeepSeek V3.2 DeepSeek | 1424.6 +82.7 | $0.290 | Balanced −23% Chat −37% Code gen −45% Agentic −9% | 262K → 164K context 236K → 147K max output |
| DeepSeek V3.2 Exp DeepSeek | 1424.4 +82.5 | $0.305 | Balanced −19% Chat −23% Code gen −40% | 262K → 164K context 236K → 66K max output |
| Qwen3 235B A22B Instruct 2507 Qwen | 1419.3 +77.4 | $0.205 | Balanced −46% Summarise −45% Chat −35% Code gen −37% Agentic −27% | 236K → 16K max output no extended reasoning |
| Qwen3 Next 80B A3B Instruct Qwen | 1418.6 +76.7 | $0.350 | Balanced −7% Summarise −31% | no extended reasoning |
| Step 3.5 Flash StepFun | 1403.8 +61.9 | $0.150 | Balanced −60% Summarise −47% Chat −58% Code gen −62% Agentic −41% | 236K → 66K max output |
| Qwen3.5-Flash Qwen | 1397.6 +55.7 | $0.114 | Balanced −70% Summarise −64% Chat −66% Code gen −69% Agentic −57% | 236K → 66K max output |
| Qwen3 30B A3B Instruct 2507 Qwen | 1384.3 +42.4 | $0.084 | Balanced −78% Summarise −73% Chat −75% Code gen −77% Agentic −68% | 236K → 32K max output no extended reasoning |
| Solar Pro 4 Upstage | 1376.2 +34.3 | $0.052 | Balanced −86% Summarise −87% Chat −86% Code gen −86% Agentic −87% | 236K → 131K max output |
| GLM 4.5 Air Z.ai | 1382.8 +40.9 | $0.310 | Balanced −18% Summarise −34% Chat −9% Code gen −5% Agentic −20% | 262K → 131K context 236K → 98K max output |
| gpt-oss-120b OpenAI | 1365.6 +23.7 | $0.070 | Balanced −81% Summarise −79% Chat −79% Code gen −80% Agentic −74% | 262K → 131K context 236K → 118K max output |
| Gemma 3 27B Google | 1358.3 +16.4 | $0.172 | Balanced −54% Summarise −58% Chat −49% Code gen −49% Agentic −49% | 236K → 118K max output no extended reasoning |
| GLM 4.7 Flash Z.ai | 1352.9 +11 | $0.145 | Balanced −62% Summarise −69% Chat −57% Code gen −56% Agentic −63% | 262K → 203K context 236K → 16K max output |
| Mercury 2 Inception | 1357.8 +15.9 | $0.375 | Balanced −1% Chat −10% Code gen −9% Agentic −13% | 262K → 128K context 236K → 50K max output |
| GPT-5.6 Luna OpenAI | 1428.5 +86.6 | $0.199 summarise | Summarise −3% | 236K → 128K max output |
| GPT-5.4 Nano OpenAI | 1372.8 +30.9 | $0.201 summarise | Summarise −2% | 236K → 128K max output |
| Qwen3 Next 80B A3B Thinking Qwen | 1367.5 +25.6 | $0.203 summarise | Summarise −2% | 236K → 33K 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/).