Alibaba Open weights: Qwen/Qwen3.8-2.4T-A95BReasoning modelCompare with another model

Qwen3.8 2.4T A95B

Reasoning effort

Qwen3.8 2.4T A95B is an Alibaba open-weights reasoning model, released 12 Aug 2026. Its best configuration ranks #70 of 775 on the BenchLeader Index at 63.2 ±5.0, in the upper half. The ± is the point: 162 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 agents & tools (78) and lowest in knowledge (64). At $3.00 per million tokens blended it is pricier than most ranked models. Output speed of 42 tokens per second puts it in the slowest quarter, with a first answer in 51.3 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 12 Aug 2026.

Blended price: $3.00/M; $2.00 in · $6.00 out
Output speed: 42 tok/s; measured by Artificial Analysis
First answer: 51 s; first token 1.00 s
Context: 1M
Full answer: 63 s; median, reasoning included
Cost per run: $2.16; one full Intelligence Index run
Released: 12 Aug 2026

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

Overall index63
Reasoning74
Agents & tools78
Knowledge64
Long context65
Composite74

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
xhigh	–	–	42 tok/s	51 s	$0.0026	Agents & tools 65
not statedbest	63.2	#70	42 tok/s	51 s	$0.0026	Agents & tools 78 · Composite 74 · Knowledge 64 · Long context 65 · Reasoning 74

How its index has moved

19 Sept 2026 to 18 Oct 2026

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.

Reasoning

Benchmark	not stated	Source
GPQA Diamond (AA)not in index	93.5%#14	Artificial Analysis
Humanity's Last Exam (AA)not in index	42.5%#67	Artificial Analysis
CritPt	20.0%#61	Artificial Analysis

Coding

Benchmark	not stated	Source
SciCode (AA)not in index	54.0%#68	Artificial Analysis
Terminal-Bench 4.0 (AA)not in index	11.1%#85	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	82.0%#39	Artificial Analysis

Agents & tools

Benchmark	xhigh	not stated	Source
MCP Atlas	84.5%#4	–	Scale AI SEAL
GDPval-AA v2.1not in indexeffort not stated	–	55.4%#31	Artificial Analysis
τ³-Banking (AA)effort not stated	–	49.1%#5	Artificial Analysis
AutomationBenchnot in indexeffort not stated	–	57.3%#48	Artificial Analysis
GDP.pdf (AA)not in indexeffort not stated	–	15.6%#93	Artificial Analysis
EnterpriseOps-Gymnot in indexeffort not stated	–	47.4%#8	Artificial Analysis

Knowledge

Benchmark	not stated	Source
AA-Omniscience	4.3#112	Artificial Analysis
AA-Omniscience: accuracynot in index	31.3%#170	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	60.8%#72	Artificial Analysis

Long context

Benchmark	not stated	Source	Trend
AA-LCR	80.3%#70	Artificial Analysis
MLCRnot in index	0.0%#90	Artificial Analysis

Composite

Benchmark	not stated	Source	Trend
AA Intelligence Index v4.3.2	40.0#62	Artificial Analysis

Safety & honesty

Benchmark	xhigh	Source	Trend
FORTRESSnot in index	21.6%#35	Scale AI SEAL

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#5τ³-Banking (AA)49.1% · of 202

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	117 tok/s	0.97 s	$2.00	$6.00	262k	–
Together	113 tok/s	3.52 s	$2.00	$6.00	1.0M	–
NovitaAI	37 tok/s	1.86 s	$2.00	$6.00	1M	–
SiliconFlow	33 tok/s	2.04 s	$2.00	$6.00	1.0M	fp8
Alibaba Cloud Int.	32 tok/s	3.24 s	$2.00	$6.00	1M	–

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

input outputnow $2.00 in · $6.00 out

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 $0.250 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0026	$0.0021	58.5 s
Summarize a 30-page report	12,000 / 600	$0.028	$0.012	1.1 min
Code edit	6,000 / 1,500	$0.021	$0.013	1.5 min
Agentic coding session	60,000 / 4,000	$0.144	$0.065	2.5 min
Structured extraction	2,000 / 200	$0.0052	$0.0026	56.1 s