Alternatives to Qwen3 235B A22B Thinking 2507

Qwen LMArena 1415.2 $0.748/M on a balanced workload prices as of 2026-10-06

16 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, 28 score at least as high as Qwen3 235B A22B Thinking 2507 and cost no more under at least one listed workload, with at least one of those dimensions improved. 16 match or beat its recorded context, maximum output, input modes, tool use and extended reasoning. 12 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

Workload	Qwen3 235B A22B Thinking 2507 $/M	Recorded capabilities preserved	Recorded capability losses
Balanced	$0.748	16	9
Summarise	$0.334	14	9
Chat	$1.06	16	10
Code gen	$1.47	16	10
Agentic	$0.541	16	11

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

Intelligence	1415.2
Context window	131K
Max output	117K
Input modes	text
Tool use	yes
Extended reasoning	yes

Preserving recorded capabilities requires matching or beating these fields. Full record for Qwen3 235B A22B Thinking 2507, or check Qwen3 235B A22B Thinking 2507 against the whole catalogue.

Recorded capabilities preserved 16

Each option has a higher score or a lower price, with neither dimension worse under the listed workloads. It matches or beats Qwen3 235B A22B Thinking 2507 on recorded context, maximum output, input modes, tool use and extended reasoning.

#	Model	Intelligence	$/M	Holds under
1	MiMo-V2.6-Pro Xiaomi	1491 +75.8	$0.540	Balanced −28% Summarise −1% Chat −55% Code gen −55% Agentic −55%
2	GLM 5.3 Flash Z.ai	1469.6 +54.4	$0.238	Balanced −68% Summarise −60% Chat −76% Code gen −76% Agentic −76%
3	DeepSeek V4.1 Flash DeepSeek	1462.5 +47.3	$0.182	Balanced −76% Summarise −80% Chat −77% Code gen −76% Agentic −81%
4	MiMo-V2.5-Pro Xiaomi	1465 +49.8	$0.544	Balanced −27% Chat −55% Code gen −55% Agentic −55%
5	MiMo-V2.6-Flash Xiaomi	1456.4 +41.2	$0.175	Balanced −77% Summarise −68% Chat −85% Code gen −86% Agentic −85%
6	Qwen3.7 Plus Qwen	1454.6 +39.4	$0.560	Balanced −25% Summarise −12% Chat −41% Code gen −41% Agentic −42%
7	Hy3 Tencent	1440.7 +25.5	$0.231	Balanced −69% Summarise −63% Chat −75% Code gen −75% Agentic −75%
8	Qwen3.8 27B Qwen	1440.7 +25.5	$0.562	Balanced −25% Summarise −45% Chat −20% Code gen −17% Agentic −31%
9	Gemma 4 26B A4B Google	1433.9 +18.7	$0.127	Balanced −83% Summarise −79% Chat −86% Code gen −86% Agentic −85%
10	DeepSeek V4 Flash 0423 DeepSeek	1432.1 +16.9	$0.113	Balanced −85% Summarise −78% Chat −90% Code gen −91% Agentic −89%
11	GLM 4.7 Z.ai	1435.5 +20.3	$0.738	Balanced −1% Chat −20% Code gen −20% Agentic −24%
12	MiniMax M3 MiniMax	1432.1 +16.9	$0.485	Balanced −35% Summarise −23% Chat −49% Code gen −49% Agentic −50%
13	MiMo-V2.5 Xiaomi	1427.5 +12.3	$0.175	Balanced −77% Summarise −68% Chat −85% Code gen −86% Agentic −85%
14	GPT-5.6 Luna OpenAI	1431 +15.8	$0.450	Balanced −40% Summarise −40% Chat −48% Code gen −47% Agentic −55%
15	DeepSeek V3.2 Exp DeepSeek	1422.4 tie	$0.305	Balanced −59% Summarise −17% Chat −69% Code gen −76% Agentic −46%
16	DeepSeek V3.1 Terminus DeepSeek	1417.1 tie	$0.453	Balanced −39% Summarise −8% Chat −47% Code gen −52% Agentic −30%

Alternatives with recorded capability losses 12

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.

Model	Intelligence	$/M	Holds under	Recorded capability loss
Gemma 4 31B Google	1443.4 +28.2	$0.205	Balanced −73% Summarise −54% Chat −77% Code gen −80% Agentic −67%	118K → 16K max output
Qwen3.6 Plus Qwen	1436.7 +21.5	$0.731	Balanced −2% Chat −8% Code gen −12%	118K → 66K max output
DeepSeek V3.2 DeepSeek	1424.5 tie	$0.290	Balanced −61% Summarise −31% Chat −75% Code gen −78% Agentic −63%	118K → 66K max output
Qwen3 235B A22B Instruct 2507 Qwen	1419.3 tie	$0.205	Balanced −73% Summarise −66% Chat −74% Code gen −75% Agentic −71%	118K → 16K max output no extended reasoning
Qwen3 VL 235B A22B Instruct Qwen	1420.1 tie	$0.370	Balanced −51% Summarise −38% Chat −58% Code gen −59% Agentic −54%	118K → 33K max output no extended reasoning
Qwen3 Next 80B A3B Instruct Qwen	1417.3 tie	$0.268	Balanced −64% Summarise −61% Chat −65% Code gen −66% Agentic −63%	no extended reasoning
DeepSeek V3.1 DeepSeek	1419.7 tie	$0.453	Balanced −39% Summarise −8% Chat −47% Code gen −52% Agentic −30%	118K → 33K max output
Gemini 3.1 Flash Lite Preview Google	1415.7 tie	$0.563	Balanced −25% Summarise −26% Chat −35% Code gen −33% Agentic −44%	118K → 66K max output
Qwen3.5-122B-A10B Qwen	1417.1 tie	$0.715	Balanced −4% Chat −7% Code gen −8% Agentic −1%	118K → 66K max output
GLM 4.6 Z.ai	1439.9 +24.7	$0.853 chat	Chat −19% Code gen −19% Agentic −22%	118K → 16K max output
Gemini 3.5 Flash Lite Google	1434.7 +19.5	$0.333 summarise	Summarise under 1% less Agentic −13%	118K → 66K max output
Gemini 2.5 Flash Google	1417 tie	$0.333 summarise	Summarise under 1% less Agentic −13%	118K → 66K 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.

Choose models

Compare costs

Build with AI

Understand the numbers

Check our work

Reuse the data