Qwen3.5 122B A10B
AlibabaReleased Feb 23, 2026Open weights · Apache 2.0qwen3.5-122b-a10b
- Reasoning
- yes
- Knowledge cutoff
- September 2025
- Context
- 262K tokens
- Max output
- 64k tokens
- Input
- text, image, video
- API features
- tools, structured outputs, JSON mode, seed, logprobs
- Alibaba's API
- yes, as qwen3.5-122b-a10b
Read Oct 23, 06:00 UTC from the hosts' published offers and OpenRouter, spec sheet from Alibaba · cutoff: aiknowledgecutoff.com
The cheapest host, SiliconFlow, serves fp8 weights (maker: bf16). Cheapest at the maker's precision and context: Novita AI, $3.60 blended. Weights and context per host
Speed by provider
Output speed of Qwen3.5 122B A10B at each host, fastest first
- Alibaba97 tok/s0.56 s to first token
- AtlasCloud85 tok/s1.92 s to first token
- SiliconFlow52 tok/s1.26 s to first token
Speed: OpenRouter, median of the last 30 min, read Oct 23, 06:00 UTC · a snapshot of real traffic before the daily read, not a daily average · bars from zero.
Run it yourself
open weights, Apache 2.0 · fits 4 of the 6 workstations in our guideSelf-host guide ›- Parameters
- 122B
- 10B active per token (MoE)
- Memory needed
- 73 GB
- in INT4 · 146 GB in INT8
- Smallest cloud setup
- A100
- 80 GB · $1.19/h, about $869 a month
- Worth it above
- 371M tokens
- a month, against the cheapest API at $2.34 per M
| Quantization | Memory | Smallest cloud GPU | Quality |
|---|---|---|---|
| INT4 (Q4) | 73 GB | A100 · $1.19/h | smallest footprint, minor quality loss |
| INT8 (Q8) | 146 GB | 2× H100 · $4.30/h | near-lossless |
| FP16 (full) | 293 GB | 4× H100 · $8.60/h | reference quality |
At home: in INT4 it fits a Mac Studio M5 Ultra (96 GB unified), about $5,499 once, or $0.2092/h over 3 years, electricity aside; it pays for itself after about 2B tokens against the cheapest API.
Memory ≈ parameters × bytes per parameter × 1.2 · cloud price: on-demand rate, running 24/7 · builds: GGUF, AWQ, MLX, GPTQ, Ollama · break-even calculator
| Provider | Input | Output | Cache ↑ | tok/s | First token | Uptime 24h | Checked |
|---|---|---|---|---|---|---|---|
| AtlasCloudfp8 (bf16) · 256K ctx · 64K out | $0.30 | $2.40 | $0.30 | 85 | 1.92 s | 99.3% | Oct 23 |
| SiliconFlowfp8 (bf16) · 256K ctx | $0.26 | $2.08 | — | 52 | 1.26 s | 97.2% | Oct 23 |
| DeepInfrafp4 (bf16) · 256K ctx | $0.29 | $2.40 | — | — | — | 100% | Oct 1 |
| Alibabaprecision not read: host terms · 256K ctx · 64K out | $0.40 | $3.20 | — | 97 | 0.56 s | 100% | Oct 23 |
| Novita AIbf16 · 256K ctx · 64K out | $0.40 | $3.20 | — | — | — | — | Oct 23 |
5 of 5 providers · prices per million tokens, read Oct 23, 06:00 UTC · speed: OpenRouter, median of the last 30 min, read Oct 23, 06:00 UTC · uptime: OpenRouter, last 24 h · under each host: weights precision, context and max output as served, the maker's value in brackets when the host's differs
| Provider | Weights | Context | Max output | Tools | Structured outputs | Reasoning |
|---|---|---|---|---|---|---|
| SiliconFlow | fp8 | 256K | not read: host terms | no | yes | yes |
| DeepInfra | fp4 | 256K | not stated by the host | — | — | — |
| AtlasCloud | fp8 | 256K | 64K | yes | yes | yes |
| Alibaba | not read: host terms | 256K | 64K | yes | yes | yes |
| Novita AI | bf16 | 256K | 64K | yes | no | yes |
Weights, context and max output: read at each host's own catalogue or docs where its terms allow, otherwise as the host declares to OpenRouter; hover a cell for its source and date · tools, structured outputs, reasoning: as declared to OpenRouter · not verified by us
Qwen3.5 122B A10B spec sheet, field by field, with sources
Qwen3.5 122B A10B spec sheetmaker-declared
What Alibaba declares about the model itself — read from its own weights repository and documentation, not from any host. What each endpoint actually serves is in the pricing above: the two can differ, and both are true.
- Context window
- 262,144 tokens
- Parameters
- 125,086,497,008≈ 125.1B
- Active experts
- 8 of 256 active per token
- Weight precision
- BF16
- License
- Apache 2.0 — open weights, permissivecurated · read 2026-09-07Scope: the licence on the published weights, and nothing else. It is not the maker's acceptable-use policy, which is a separate document, and it is not the contract you sign with whichever host you call — hosts set their own terms, which this index does not read.
- Knowledge cutoff
- not published by the maker≈ September 2025 third-party estimate — not the maker's figurecurated
- Modalities
- text · image · video
- Max output (official)
- to verify
- Release date
- to verify≈ February 23, 2026 third-party catalogue — not the maker's figure
- Maker lifecycle status
- to verify
Also declared in the weights
- Architecture class
- Qwen3_5MoeForConditionalGeneration
- Layers
- 48
- Vocabulary
- 248,320 tokens
- Weights first published
- February 24, 2026
- Weights last updated
- April 24, 2026
Qwen/Qwen3.5-122B-A10B · config.json @ dc4d348 · read 2026-10-08 · 6 of 10 fields published by the maker
| Provider | Chatbot 1,000 in / 500 out | RAG / long context 8,000 in / 500 out | Batch summarize 4,000 in / 1,000 out |
|---|---|---|---|
| SiliconFlow | $1.30 | $3.12 | $3.12 |
| DeepInfra | $1.49 | $3.52 | $3.56 |
| AtlasCloud | $1.50 | $3.60 | $3.60 |
| Alibaba | $2.00 | $4.80 | $4.80 |
| Novita AI | $2.00 | $4.80 | $4.80 |
What each host bills for 1,000 requests of a representative job
Your own workload
Other Alibaba models · all its models, hosts and history ›
- Qwen3.8 27B16 providers · from $1.29
- Qwen3.6 35B A3B12 providers · from $1.00
- Qwen3 235B A22B Instruct 250710 providers · from $0.44
- Qwen3.5 397B A17B10 providers · from $3.45
- Qwen3.6 27B9 providers · from $2.25
- Qwen3.5-35B-A3B8 providers · from $1.14