Today's price · per 1M tokens
Input
$0.400
per 1M tokens
Output
$3.20
per 1M tokens
Blended
$1.10
blended $/1M — 3:1 weighted input:output
How this price is scoped: Alibaba Model Studio International list price, single 0<Token≤256K tier, from the vendor's open-source Qwen table. Thinking-mode output is billed at the same rate as non-thinking. No context-cache or batch rate is printed for this SKU, so cached input is left unset rather than assumed. Gateway resellers quote lower routed rates ($0.29/$2.40). This SKU's row carries no discount label, so the standard list rate is the one tracked here — the same basis as every other Alibaba row on this site.
Price receipt
We read Alibaba's own pricing page on and found qwen3.5-122b-a10b listed with a price on the same row — that read is where the number above comes from, recorded as a new model. We keep the page text we read; its content hash is ff4af67ce0.
Overview
Open-weights Qwen3.5 MoE (122B total, 10B active) served by Alibaba, the model's maker, on Model Studio. $0.40/$3.20 per 1M tokens, a single 0–256K tier, with thinking-mode output billed at the same rate as non-thinking. 256K-token context, text-only.
Deploy this open model on rented GPUs
Open weights mean you can self-host instead of paying per-token API prices. These platforms let you serve it on demand — often cheaper at scale.
Partner links — we may earn a commission if you sign up, at no cost to you. We only list platforms we'd recommend regardless.
Capabilities
struck through = not supportedBenchmark performance
accuracy % · higher is betterWe hold no per-benchmark accuracy scores for this model yet, so it has no accuracy average. That is a gap in our coverage, not a sign the model is untested — independent evaluators often publish a composite index for a new model long before releasing its per-benchmark numbers. The percentile above is its standing across the independent composites below.
- GDPval-AA v2: 821.1 ELO (Real-world work tasks, human baseline = 1000)
- Arena (arena.ai): 1416.4 ELO (Human preference)
- Arena WebDev: 1360.2 ELO (Human preference on building web apps)
- Arena Vision: 1225.8 ELO (Human preference on image-input prompts)
- Humanity's Last Exam: 25.2% (reasoning effort)
- GPQA Diamond: 85.7% (reasoning effort)
- AA-LCR (long-context reasoning): 76.3% (reasoning effort)
- CritPt (research-level physics): 0.6% (reasoning effort)
- IFBench (instruction following): 75.7% (reasoning effort)
- τ²-Bench Telecom: 93.6% (reasoning effort)
- Terminal-Bench Hard: 31.1% (reasoning effort)
- Terminal-Bench 2.1: 47.6% (reasoning effort)
- Terminal-Bench 4.0: 0% (reasoning effort)
- SciCode: 39.7% (reasoning effort)
- MMMU-Pro: 75% (reasoning effort)
- Measured by AA: 108.1 tokens/sec output, 4.47s to first token, 18.84s to first answer token (reasoning effort)
- Ranks #128 of 189 comparably-measured models by percentile score, across 2 independent measurements
- Ranks #109 of 189 comparably-tested models by normalized performance per dollar
108.1 tokens/sec output 4.47s latency to first token (TTFT)
Throughput: Artificial Analysis, measured 2026-11-14 · Source: Artificial Analysis (artificialanalysis.ai)
Source: Artificial Analysis, LMSYS Chatbot Arena (UC Berkeley) · updated · See full rankings →
Specifications
- Provider
- Alibaba
- Context window
- 262K tokens
- Modality
- text
- Parameters
- 122B (A10B)
- Open source
- Yes — open weights available
- Released
- Status
- Current
- Last updated
- Tags
Availability verified: — listed on Alibaba's own page