What are Qwen3.5-35B-A3B's benchmark scores?

As of 25 Oct 2026, Qwen3.5-35B-A3B (thinking effort) scores 52.8 ±7.4 on the BenchLeader Index, strongest in instruction following (71) and weakest in knowledge (40). That rests on 7 independent benchmark results, and every score on this page links to its publisher.

How does Qwen3.5-35B-A3B rank?

Qwen3.5-35B-A3B (thinking effort) is #304 of 785 ranked configurations on 25 Oct 2026. 406 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.

How much does Qwen3.5-35B-A3B cost per 1M tokens?

As of 25 Oct 2026, Qwen3.5-35B-A3B costs $0.250 per 1M input tokens and $2.00 per 1M output tokens, $0.688 blended (three input tokens per output token).

When was Qwen3.5-35B-A3B released?

Alibaba released Qwen3.5-35B-A3B on 24 Feb 2026. BenchLeader has results for it at 3 reasoning-effort settings.

Alibaba Open weights: Qwen/Qwen3.5-35B-A3B Compare with another model

Qwen3.5-35B-A3B

Reasoning effort

Qwen3.5-35B-A3B is an Alibaba open-weights model, released 24 Feb 2026. Its best configuration (thinking reasoning effort) ranks #304 of 785 on the BenchLeader Index at 52.8 ±7.4, in the upper half. The ± is the point: 406 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (71) and lowest in knowledge (40). At $0.688 per million tokens blended it is mid-priced. Output speed of 148 tokens per second puts it in the fastest quarter, with a first answer in 15.5 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 24 Feb 2026.stale: no new result in six months

Blended price: $0.688/M; $0.250 in · $2.00 out
Output speed: 148 tok/s; measured by Artificial Analysis
First answer: 15 s; first token 0.94 s
Context: 262k
Full answer: 19 s; median, reasoning included
Cost per run: –
Released: 24 Feb 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index53
Reasoning42
Agents & tools57
Knowledge40
Instruction following71
Multimodal56
Long context61
Composite51

Versions

Alibaba has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
Qwen3.6 35B A3Bthinking	14 Apr 2026	52.2	#326
Qwen3.5-35B-A3Bthinkingthis page	24 Feb 2026	52.8	#304

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
no reasoning	45.6	#510	169 tok/s	1.99 s	$0.0007	Agents & tools 36 · Coding 25 · Composite 46 · Instruction following 47 · Knowledge 33 · Long context 56 · Maths 54 · Multimodal 52 · Reasoning 50
thinkingbest	52.8	#304	148 tok/s	15 s	$0.0007	Agents & tools 57 · Composite 51 · Instruction following 71 · Knowledge 40 · Long context 61 · Multimodal 56 · Reasoning 42
not stated	50.2	#377	68 tok/s	0.94 s	$0.0007	Coding 38 · Human preference 57 · Maths 48 · Reasoning 59

How its index has moved

8 Sept 2026 to 23 Oct 2026

thinking
not stated
no reasoning

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Metadata

Benchmark	thinking	Source	Trend
AA Openness Indexnot in index	38.9#34	Artificial Analysis

Reasoning

Benchmark	no reasoning	thinking	not stated	Source
GPQA Diamond	81.2%#121	–	83.5%#101	Epoch AI Benchmarking Hub
Arena Hard Promptseffort not stated	–	–	1412#162	Arena (LMArena)
GPQA Diamond (AA)not in index	81.9%#176	84.5%#137	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	13.4%#270	21.0%#211	–	Artificial Analysis
CritPt	0.6%#254	0.9%#237	–	Artificial Analysis
LMCAnot in indexeffort not stated	–	–	29.5%#138	Epoch AI Benchmarking Hub
DTBenchnot in indexeffort not stated	–	–	80.0%#132	Epoch AI Benchmarking Hub

Coding

Benchmark	no reasoning	not stated	Source
SciCode	29.3%#199	–	SciCode
Arena Codingeffort not stated	–	1434#164	Arena (LMArena)
Arena WebDeveffort not stated	–	1254#126	Arena (LMArena)
Terminal-Bench 2.1 (AA)not in index	40.8%#137	–	Artificial Analysis

Agents & tools

Benchmark	no reasoning	thinking	Source
Terminal-Bench Hard	10.6%#221	26.5%#129	Artificial Analysis
τ²-Bench Telecom (AA)not in index	86.3%#88	89.2%#69	Artificial Analysis
τ³-Banking (AA)	5.0%#185	–	Artificial Analysis
ITBench SRE (AA)not in index	–	21.5%#39	Artificial Analysis

Maths

Benchmark	no reasoning	not stated	Source
OTIS Mock AIME	70.0%#140	54.4%#177	Epoch AI Benchmarking Hub
AIME 2026effort not statedpossible contamination	–	93.3%#22	MathArena
HMMT February 2026effort not statedpossible contamination	–	81.8%#25	MathArena
MathArena Apexeffort not statedpossible contamination	–	4.2%#28	MathArena
Arena Mathseffort not stated	–	1398#152	Arena (LMArena)

Knowledge

Benchmark	no reasoning	thinking	Source
AA-Omniscience	-63.0#485	-48.1#371	Artificial Analysis
AA-Omniscience: accuracynot in index	15.7%#404	20.1%#305	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	6.6%#503	14.6%#346	Artificial Analysis
Global-MMLU-Lite (AA)not in index	–	86.4%#25	Artificial Analysis

Instruction following

Benchmark	no reasoning	thinking	not stated	Source	Trend
IFBench	44.5%#212	72.5%#47	–	Artificial Analysis
Arena Instruction Followingnot in indexeffort not stated	–	–	1387#152	Arena (LMArena)

Human preference

Benchmark	not stated	Source
Arena Text	1394#157	Arena (LMArena)
Arena Creative Writingnot in index	1343#176	Arena (LMArena)
Arena Multi-turnnot in index	1394#161	Arena (LMArena)
Arena Longer Queriesnot in index	1399#163	Arena (LMArena)

Multimodal

Benchmark	no reasoning	thinking	Source	Trend
MMMU-Pro	69.2%#156	72.7%#131	Artificial Analysis

Long context

Benchmark	no reasoning	thinking	Source	Trend
AA-LCR	63.0%#268	72.0%#191	Artificial Analysis

Composite

Benchmark	no reasoning	thinking	not stated	Source	Trend
Epoch Capabilities Indexnot in indexeffort not stated	–	–	142.5#100	Epoch AI Benchmarking Hub
AA Intelligence Index v4.3.2	15.1#288	19.3#244	–	Artificial Analysis

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Venice	186 tok/s	0.77 s	$0.150	$1.00	256k	–
Parasail	108 tok/s	0.94 s	$0.150	$1.00	262k	fp8
DeepInfra	97 tok/s	43 s	$0.140	$1.00	262k	fp8
Alibaba Cloud Int.	73 tok/s	1.29 s	$0.163	$1.30	262k	–
AtlasCloud	60 tok/s	1.97 s	$0.225	$1.80	262k	fp8
SiliconFlow	42 tok/s	7.32 s	$0.240	$1.80	262k	fp8
Darkbloom	25 tok/s	4.14 s	$0.065	$0.750	262k	fp4

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.150 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0007	$0.0007	17.5 s
Summarize a 30-page report	12,000 / 600	$0.0042	$0.0033	19.5 s
Code edit	6,000 / 1,500	$0.0045	$0.0041	25.6 s
Agentic coding session	60,000 / 4,000	$0.023	$0.018	42.4 s
Structured extraction	2,000 / 200	$0.0009	$0.0008	16.8 s

New this month

Questions people ask

What are Qwen3.5-35B-A3B's benchmark scores?: As of 25 Oct 2026, Qwen3.5-35B-A3B (thinking effort) scores 52.8 ±7.4 on the BenchLeader Index, strongest in instruction following (71) and weakest in knowledge (40). That rests on 7 independent benchmark results, and every score on this page links to its publisher.
How does Qwen3.5-35B-A3B rank?: Qwen3.5-35B-A3B (thinking effort) is #304 of 785 ranked configurations on 25 Oct 2026. 406 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
How much does Qwen3.5-35B-A3B cost per 1M tokens?: As of 25 Oct 2026, Qwen3.5-35B-A3B costs $0.250 per 1M input tokens and $2.00 per 1M output tokens, $0.688 blended (three input tokens per output token).
When was Qwen3.5-35B-A3B released?: Alibaba released Qwen3.5-35B-A3B on 24 Feb 2026. BenchLeader has results for it at 3 reasoning-effort settings.

Data as of 25 Oct 2026. Compare these configurations.

Cite as: BenchLeader, “Qwen3.5-35B-A3B: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/qwen3-5-35b-a3b.