ai.canshu.com
中文
VRAM Calculator Advisor & TCO API Pricing Compare Models Leaderboard GPUs

Search models, GPUs, and tools

↑↓ Navigate · Enter Open · ★ Watchlist prioritized ai.canshu.com
Alibaba Qwen
Open-Weight Alibaba Qwen · MoE 122B / A10B · Text+Image+Video · BF16

Qwen3.5-122B-A10B (MoE) Architecture Parameters, Benchmarks & VRAM

Sweet-spot 122B MoE (10B active) for 128GB unified-memory workstations (~75GB at Q4_K_M).

🧬 Qwen3.5-122B-A10B (MoE) Architecture & Tensor Specifications

Verified directly against HuggingFace (Qwen/Qwen3.5-122B-A10B) config.json & Model Card

HF Class: Qwen3_5MoeForConditionalGeneration

1. Backbone & Dimensions

qwen3_5_moe
Total / Active Params
122B / 10B
Hidden Size (d_model)
3,072
MoE Expert Dim
1,024
MoE Experts
256 Routed · Top-8 + 1 Shared
Activation / Norm
SiLU (SwiGLU) · ε=1e-6
Native Tensor Dtype
BF16

2. Attention & Layer Layout

HYBRID
Num Hidden Layers
48 Layers (12 Global KV)
Hidden Layer Layout
48 Hybrid MoE Layers: 12 × [3 × Gated DeltaNet→MoE + 1 × Gated Attention→MoE]
Attention Heads (Q / KV)
32 Q / 2 KV (16:1 GQA)
Head Dimension
256
Linear / KV Compression
Only 12 layers × 2 KV heads cached; 256 routed experts (8 active)
Vision Encoder / MTP
ViT 27L · MTP 1L

3. Context, RoPE & Tokenizer

Tokenizer
Native Context Length
262,144 tokens
Max Extended Context
1,010,000 tokens
RoPE Configuration
mRoPE (θ=10,000,000 · partial=0.25 · YaRN 4x→1M)
Vocab Size
248,320 (Untied)
Special Token IDs
BOS: 248044 · EOS: [248046,248044] · PAD: 248044

📋 Qwen3.5-122B-A10B (MoE) Official API & System Card Specifications

Verified against official Alibaba Qwen developer documentation & pricing

API Model ID: Qwen/Qwen3.5-122B-A10B

1. Identity & Capabilities

API Spec
API Model ID
Qwen/Qwen3.5-122B-A10B
Release Date
2026-02-23
Weight License
Open-Weight (API Spec)
Native Reasoning
✅ Supported
Function Calling & JSON Schema
✅ 原生支持
Batch API 异步批处理
✅ Supported (50% off)

2. Context, Output & Modalities

Limits
Max Context Window
256,000 tokens (256K)
Max Output Tokens
66,000 tokens
Input Modalities
Text + Image + Video
Output Modalities
Text
Direct China Access
✅ Direct CN + RMB
Deployment Mode
Open-Weight + Cloud API

3. Official Pricing & Cache

CNY / 1M
Input / 1M Tokens
¥ 0.8
Cached Input / 1M
¥ 0.8
Output / 1M Tokens
¥ 6.38
Tiered Pricing Notes
SiliconFlow: Open-weight cloud hosting on SiliconFlow

🏆 Official Benchmark Results

Sourced from official System Card / Model Card & technical reports

Arena Elo ~1478 AA Index 48
📚 Knowledge
MMLU-Pro 86.7 %

A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks.

📚 Knowledge
C-Eval 91.9 %

C-Eval is a comprehensive Chinese evaluation suite designed to assess advanced knowledge and reasoning abilities of foundation models in a Chinese context.

🧠 Reasoning
GPQA Diamond 86.6 %

PhD-level Google-proof science reasoning questions

💻 Coding
SWE-bench Verified 72 %

A verified subset of 500 software engineering problems from real GitHub issues, validated by human annotators for evaluating language models' ability to resolve real-world coding issues by generating patches for Python codebases.

💻 Coding
LiveCodeBench v6 78.9 %

LiveCodeBench is a holistic and contamination-free evaluation benchmark for large language models for code.

🤖 Agentic
BFCL-V4 72.2 %

Berkeley Function Calling Leaderboard V4 (BFCL-V4) evaluates LLMs on their ability to accurately call functions and APIs, including simple, multiple, parallel, and nested function calls across diverse programming scenarios.

👁️ Multimodal
MMMU 83.9 %

MMMU (Massive Multi-discipline Multimodal Understanding) is a benchmark designed to evaluate multimodal models on college-level subject knowledge and deliberate reasoning.

🎛️ Official Sampling & Reasoning Parameters

Recommended generation settings & API parameter constraints

Thinking Mode
temperature=0.6 top_p=0.95 top_k=20 max_tokens=32,768

📊 Qwen3.5-122B-A10B (MoE) Full Benchmark Scorecard

81 public benchmarks · ranks among 154 models · synced 2026-11-20

⚔️ Compare head-to-head

🤖 Agents & Tools 11 items

t2-bench #8/12 79.5%

t2-bench is a benchmark for evaluating agentic tool use capabilities, measuring how well models can select, sequence, and utilize tools to solve complex tasks. It tests autonomous planning and execution in multi-step scenarios.

BFCL-V4 #5/10 72.2%

Berkeley Function Calling Leaderboard V4 (BFCL-V4) evaluates LLMs on their ability to accurately call functions and APIs, including simple, multiple, parallel, and nested function calls across diverse programming scenarios.

BrowseComp-zh #3/7 69.9%

A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11 diverse domains including Film & TV, Technology, Medicine, and History.

AndroidWorld_SR #2/3 66.4%

AndroidWorld Success Rate (SR) benchmark - A dynamic benchmarking environment for autonomous agents operating on Android devices.

BrowseComp #38/54 63.8%

BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information.

WideSearch #10/13 60.5%

WideSearch is an agentic search benchmark that evaluates models' ability to perform broad, parallel search operations across multiple sources. It tests wide-coverage information retrieval and synthesis capabilities.

OSWorld-Verified #22/26 58%

OSWorld-Verified is a verified subset of OSWorld, a scalable real computer environment for multimodal agents supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS.

TIR-Bench #4/4 53.2%

A tool-calling and multimodal interaction benchmark for testing visual instruction following and execution reliability.

Seal-0 #4/5 44.1%

Seal-0 is a benchmark for evaluating agentic search capabilities, testing models' ability to navigate and retrieve information using tools.

VITA-Bench #7/8 33.6%

VITA-Bench evaluates AI agents on real-world virtual task automation, measuring their ability to complete complex multi-step tasks in simulated environments.

DeepPlanning #5/7 24.1%

DeepPlanning evaluates LLMs on complex multi-step planning tasks requiring long-horizon reasoning, goal decomposition, and strategic decision-making.

📐 Math 13 items

MMLU-Redux #5/15 94%

An improved version of the MMLU benchmark featuring manually re-annotated questions to identify and correct errors in the original dataset. Provides more reliable evaluation metrics for language models by addressing dataset quality issues found in the original MMLU.

HMMT 2025 #9/11 91.4%

Harvard-MIT Mathematics Tournament 2025 - A prestigious student-organized mathematics competition for high school students featuring two tournaments (November 2025 at MIT and February 2026 at Harvard) with individual tests, team rounds, and guts rounds.

HMMT25 #4/9 90.3%

Harvard-MIT Mathematics Tournament 2025 - A prestigious student-organized mathematics competition for high school students featuring two tournaments (November 2025 at MIT and February 2026 at Harvard) with individual tests, team rounds, and guts rounds.

MathVista-Mini #3/7 87.4%

MathVista-Mini is a smaller version of the MathVista benchmark that evaluates mathematical reasoning in visual contexts.

MMLU-Pro #9/39 86.7%

A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks.

MMMLU #12/20 86.7%

Multilingual Massive Multitask Language Understanding dataset released by OpenAI, featuring professionally translated MMLU test questions across 14 languages including Arabic, Bengali, German, Spanish, French, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese, Swahili…

MathVision #10/20 86.2%

MATH-Vision is a dataset designed to measure multimodal mathematical reasoning capabilities.

DynaMath #3/8 85.9%

A multimodal mathematics and reasoning benchmark focused on dynamic visual problem solving.

MMLU-ProX #6/13 82.2%

Extended version of MMLU-Pro providing additional challenging multiple-choice questions for evaluating language models across diverse academic and professional domains. Built on the foundation of the Massive Multitask Language Understanding benchmark framework.

PolyMATH #6/9 68.9%

Polymath is a challenging multi-modal mathematical reasoning benchmark designed to evaluate the general cognitive reasoning abilities of Multi-modal Large Language Models (MLLMs).

SuperGPQA #6/16 67.1%

SuperGPQA is a comprehensive benchmark that evaluates large language models across 285 graduate-level academic disciplines.

Humanity's Last Exam #25/69 47.5%

Humanity's Last Exam (HLE) is a multi-modal academic benchmark with 2,500 questions across mathematics, humanities, and natural sciences, designed to test LLM capabilities at the frontier of human knowledge with unambiguous, verifiable solutions.

CodeForces #5/8 0.9

A competitive programming benchmark using problems from the CodeForces platform. The benchmark evaluates code generation capabilities of LLMs on algorithmic problems with difficulty ratings ranging from 800 to 2400.

👁️ Multimodal 39 items

VLMsAreBlind #4/6 96.7%

A vision-language benchmark that probes blind spots and brittle reasoning in multimodal models.

AI2D #2/7 93.3%

AI2D is a dataset of 4,903 illustrative diagrams from grade school natural sciences (such as food webs, human physiology, and life cycles) with over 15,000 multiple choice questions and answers.

V* #5/7 93.2%

A visual reasoning benchmark evaluating multimodal inference under challenging spatial and grounded tasks.

MMBench-V1.1 #1/6 92.8%

Version 1.1 of MMBench, an improved bilingual benchmark for assessing multi-modal capabilities of vision-language models through multiple-choice questions in both English and Chinese, providing systematic evaluation across diverse vision-language tasks.

OCRBench #2/5 92.1%

OCRBench: Comprehensive evaluation benchmark for assessing Optical Character Recognition (OCR) capabilities in Large Multimodal Models across text recognition, scene text VQA, and document understanding tasks.

OmniDocBench 1.5 #6/16 89.8%

OmniDocBench 1.5 is a comprehensive benchmark for evaluating multimodal large language models on document understanding tasks, including OCR, document parsing, information extraction, and visual question answering across diverse document types.

VideoMME w sub. #3/8 87.3%

The first-ever comprehensive evaluation benchmark of Multi-modal LLMs in Video analysis. Features 900 videos (254 hours) with 2,700 question-answer pairs covering 6 primary visual domains and 30 subfields.

MLVU #2/7 87.3%

A comprehensive benchmark for multi-task long video understanding that evaluates multimodal large language models on videos ranging from 3 minutes to 2 hours across 9 distinct tasks including reasoning, captioning, recognition, and summarization.

RealWorldQA #8/13 85.1%

RealWorldQA is a benchmark designed to evaluate basic real-world spatial understanding capabilities of multimodal models.

VideoMME w/o sub. #1/4 83.9%

Video-MME is a comprehensive evaluation benchmark for multi-modal large language models in video analysis. It features 900 videos across 6 primary visual domains with 30 subfields, ranging from 11 seconds to 1 hour in duration, with 2,700 question-answer pairs.

MMMU #2/16 83.9%

MMMU (Massive Multi-discipline Multimodal Understanding) is a benchmark designed to evaluate multimodal models on college-level subject knowledge and deliberate reasoning.

MMStar #2/8 82.9%

MMStar is an elite vision-indispensable multimodal benchmark comprising 1,500 challenge samples meticulously selected by humans to evaluate 6 core capabilities and 18 detailed axes.

VideoMMMU #11/14 82%

Video-MMMU evaluates Large Multimodal Models' ability to acquire knowledge from expert-level professional videos across six disciplines through three cognitive stages: perception, comprehension, and adaptation.

CC-OCR #3/6 81.8%

A comprehensive OCR benchmark for evaluating Large Multimodal Models (LMMs) in literacy. Comprises four OCR-centric tracks: multi-scene text reading, multilingual text reading, document parsing, and key information extraction.

SlakeVQA #1/3 81.6%

A semantically-labeled knowledge-enhanced dataset for medical visual question answering. Contains 642 radiology images (CT scans, MRI scans, X-rays) covering five body parts and 14,028 bilingual English-Chinese question-answer pairs annotated by experienced physicians.

LingoQA #3/4 80.8%

A benchmark for multimodal spatial-language understanding and visual-linguistic question answering.

CharXiv-R #34/41 77.2%

CharXiv-R is the reasoning component of the CharXiv benchmark, focusing on complex reasoning questions that require synthesizing information across visual chart elements.

MMMU-Pro #21/44 76.9%

A more robust multi-discipline multimodal understanding benchmark that enhances MMMU through a three-step process: filtering text-only answerable questions, augmenting candidate options, and introducing vision-only input settings.

MVBench #2/7 76.6%

A comprehensive multi-modal video understanding benchmark covering 20 challenging video tasks that require temporal understanding beyond single-frame analysis.

MMVU #3/7 74.7%

MMVU (Multimodal Multi-disciplinary Video Understanding) is a benchmark for evaluating multimodal models on video understanding tasks across multiple disciplines, testing comprehension and reasoning capabilities on video content.

LVBench #10/15 74.4%

LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets.

ScreenSpot Pro #6/10 70.4%

ScreenSpot-Pro is a novel GUI grounding benchmark designed to rigorously evaluate the grounding capabilities of multimodal large language models (MLLMs) in professional high-resolution computing environments.

Hallusion Bench #4/6 67.6%

A comprehensive benchmark designed to evaluate image-context reasoning in large visual-language models (LVLMs) by challenging models with 346 images and 1,129 carefully crafted questions to assess language hallucination and visual illusion.

MedXpertQA #1/6 67.3%

A comprehensive benchmark to evaluate expert-level medical knowledge and advanced reasoning, featuring 4,460 questions spanning 17 specialties and 11 body systems.

PMC-VQA #1/3 63.3%

A medical visual question answering benchmark built on biomedical literature and medical figures.

ERQA #10/13 62%

Embodied Reasoning Question Answering benchmark consisting of 400 multiple-choice visual questions across spatial reasoning, trajectory reasoning, action reasoning, state estimation, and multi-view reasoning for evaluating AI capabilities in physical world interactions.

ODinW #4/7 44.5%

Object Detection in the Wild (ODinW) benchmark for evaluating object detection models' task-level transfer ability across diverse real-world datasets in terms of prediction accuracy and adaptation efficiency.

BabyVision #13/15 40.2%

A benchmark for early-stage visual reasoning and perception on child-like vision tasks.

Nuscene #1/3 15.4%

A multimodal benchmark for scene understanding and reasoning over the nuScenes autonomous driving domain.

CountBench #6/7 1

CountBench evaluates object counting capabilities in visual understanding.

RefCOCO-avg #4/6 0.9

RefCOCO-avg measures object grounding accuracy averaged across RefCOCO, RefCOCO+, and RefCOCOg benchmarks.

EmbSpatialBench #4/6 0.8

EmbSpatialBench evaluates embodied spatial understanding and reasoning capabilities.

RefSpatialBench #3/7 0.7

RefSpatialBench evaluates spatial reference understanding and grounding.

SimpleVQA #9/13 0.6

SimpleVQA is a visual question answering benchmark focused on simple queries.

MMLongBench-Doc #5/6 0.6

MMLongBench-Doc evaluates long document understanding capabilities in vision-language models.

SUNRGBD #1/3 0.4

SUNRGBD evaluates RGB-D scene understanding and 3D grounding capabilities.

ZEROBench-Sub #1/5 0.4

ZEROBench-Sub is a subset of the ZEROBench benchmark.

Hypersim #3/3 0.1

Hypersim evaluates 3D grounding and depth understanding in synthetic indoor scenes.

ZEROBench #11/13 0.1

ZEROBench is a challenging vision benchmark designed to test models on zero-shot visual understanding tasks.

Source: ZeroEval / llm-stats.com (✓ = third-party verified, others vendor-reported).

⚡ Qwen3.5-122B-A10B (MoE) Interactive VRAM & Token Speed Calculator

Concurrency (Batch):
MoE 122B / A10B·48L (d=3072)·32Q/2KV·Ctx 256K·Text+Image+Video
Runnable Models →
128 GB Unified·614 GB/s·Safe Usable 108.8 GB
Smooth Fit77.26 / 108.8 GB~64.1 tok/s
3. Weight Quantization
4. Per-Request Context16K (16,384 tok)
Verified: 2026-11-20·Roofline bandwidth/compute estimate (Batch=1)
Smooth Fit
77.26/ 108.8 GB usable

Physical 128 GB · Load 71%

Theoretical Decode Speed
~64.1tokens/s

MoE: reads only 6.06 GB active weights/step

Max Context at Batch=1
256K/ 256K

Enable Q8/Q4 KV quantization to scale batch concurrency

Weights: 73.96 GBKV Cache (1×16K): 0.52 GBOverhead: 2.78 GB
Reserves ~15% unified memory for OS
🚀 Blazing Fast (Instant CoT)~7.8s for a 500-token response
Reference Launch CLI:
OLLAMA_CONTEXT_LENGTH=16384 ollama run qwen3.5:122b-a10b

Mac Studio M5 Max 128G Fits Qwen3.5-122B-A10B (MoE) · Cost & TCO

Approx. $3,499 (~$27/GB VRAM) · 120W TDP

📊 Qwen3.5-122B-A10B (MoE) Quantization VRAM Breakdown (16K Context)

Quantization Bits/Weight Weights KV Cache (16K) Total VRAM Quality Note
FP16 / BF16 16 bpw 244 GB 0.52 GB 250.7 GB Lossless Full Precision
FP8 / Q8_0 8.5 bpw 129.63 GB 0.52 GB 134.04 GB 99.9% Near-Lossless
GGUF Q6_K 6.56 bpw 100.04 GB 0.52 GB 103.86 GB Ultra High Quality
GGUF Q5_K_M 5.69 bpw 86.77 GB 0.52 GB 90.33 GB High Balanced Quality
Q4_K_M / AWQ 4.85 bpw 73.96 GB 0.52 GB 77.26 GB Sweet Spot (Recommended)
GGUF IQ4_XS 4.25 bpw 64.81 GB 0.52 GB 67.93 GB Compact 4-bit
GGUF Q3_K_M 3.91 bpw 59.63 GB 0.52 GB 62.64 GB Moderate Perplexity Loss
GGUF Q2_K 3.15 bpw 48.04 GB 0.52 GB 50.82 GB Extreme Compression

🖥️ GPU & Apple Silicon Compatibility for Qwen3.5-122B-A10B (MoE) (Q4_K_M)

Frequently Asked Questions

What are the exact architecture parameters of Qwen3.5-122B-A10B (MoE)?

Qwen3.5-122B-A10B (MoE) uses Qwen3_5MoeForConditionalGeneration with 48 layers (48 Hybrid MoE Layers: 12 × [3 × Gated DeltaNet→MoE + 1 × Gated Attention→MoE]), hidden_size=3072, 32 Q heads / 2 KV heads (head_dim=256), and vocab_size=248,320.

How much VRAM does Qwen3.5-122B-A10B (MoE) require?

At 4-bit quantization (Q4_K_M, 16K context), Qwen3.5-122B-A10B (MoE) requires 77.26 GB VRAM. Full FP16 requires 250.7 GB.