Qwen3-235B: the model previously served by Gonka

Qwen3-235B-A22B-Instruct-2507-FP8 is a Large Language Model (LLM) from the Qwen3 family, developed by the Qwen team at Alibaba Cloud. The full name deciphers as: Qwen3 — the third generation of the series, 235B — 235 billion parameters in total, A22B — 22 billion active parameters per query, Instruct — a version trained to follow instructions, 2507 — July 2025 release, FP8 — 8-bit quantization for memory optimization.

A key architectural feature is MoE (Mixture of Experts). Unlike "dense" models (GPT-5.5, Claude Sonnet 4.6), where each token passes through all parameters, an MoE model activates only a subset of "experts" — specialized blocks of the neural network — for each query. In the case of Qwen3-235B, out of 235 billion parameters, only 22 billion are activated for each token — less than 10%. This delivers quality comparable to models with 200B+ parameters at the computational cost of a 22B model.

Practically, this means: the model is smarter than one might expect from its speed. It processes queries significantly faster than dense models of comparable quality, while requiring orders of magnitude less VRAM for inference. This is why MoE has become the dominant architecture for the largest models of 2025–2026.

The context window for Qwen3-235B is 131,072 tokens (~100,000 words) — sufficient for analyzing entire books, codebases, or long legal documents in a single query. The model supports 119 languages, including Russian, English, Chinese, Arabic, Hindi, and dozens of others — making it one of the most multilingual models on the market.

Qwen3-235B competes with the largest closed and open models. Here's a comparison of key characteristics:

Qwen3-235B delivers quality comparable to GPT-5.5 and Claude Sonnet 4.6 on most benchmarks, and as an open-weights MoE model it costs many times cheaper than proprietary alternatives: the MoE architecture activates only a portion of parameters per request and dramatically reduces computational costs. The Gonka network applies that same principle of cheap decentralized inference today to its live models — MiniMax M2.7, DeepSeek V4 Flash and GLM-5.3 Flash, available through the JoinGonka Gateway at $0.0062 per 1M tokens (hundreds to thousands of times cheaper than GPT-5.5 and Claude).

On MMLU-Pro, HumanEval, MATH-500 and GSM8K benchmarks, the model ranks among the top three open-source models, trailing only DeepSeek-R1 in mathematical reasoning tasks. In code generation, translation, and instruction-following tasks, Qwen3-235B consistently outperforms Llama 4 Maverick and matches Claude Sonnet 4.6.

When Qwen3-235B was the network's primary model, it ran on the Gonka network in a distributed fashion — via the DiLoCo protocol adapted for inference. The full model in FP8 format requires around 640 GB of video memory (VRAM), which cannot fit on a single GPU — even an H100 80GB or H200 141GB is not enough. That's why the model was split by layers (tensor parallelism + pipeline parallelism) across multiple ML nodes.

In practice, Qwen3-235B ran on a cluster of 8–16 GPU nodes, each with at least 40 GB of VRAM. Transfer Agents routed the request to the right cluster, vLLM on each node processed its fragment of the model, and the results were aggregated and returned to the user. The whole process took hundreds of milliseconds — the user never felt that their request was handled by a dozen GPUs at different points on the planet. The network applies that same principle of distributed inference today — now to its live models.

An important technical detail: Gonka uses vLLM as its serving engine. vLLM is an open-source project that delivers high-performance text generation through PagedAttention — an algorithm that optimizes video memory usage when processing many requests in parallel. This lets each node handle many requests simultaneously without losing quality.

The model supports native tool calling — invoking functions and tools directly from the model's response. This capability was added to Gonka via PR #767 with a threshold of 0.958 for detecting tool calls. On Qwen3-235B, this let developers build AI agents that interact with external APIs, databases, and tools — all through a single request. Tool calling support lives on in the network's live models as well.

Today the Gonka network has 1,293 GPUs (H100, H200, B300 and others). Where that power once went to Qwen3-235B, today it serves the network's live models — MiniMax M2.7, DeepSeek V4 Flash and GLM-5.3 Flash.

Short answer: through the Gonka network — not anymore. Qwen3-235B has been removed from the network, so you can no longer call it by the identifier qwen3-235b-a22b through our Gateway. Since the model is open (Apache 2.0), you can still run it yourself or access it through third-party open-weights hosting providers — for example, via OpenRouter (roughly $0.071/$0.100 per 1M tokens).

But if what you need is that same cheap decentralized inference people come to Gonka for — it hasn't gone anywhere, it just now runs on the network's active models. Through the JoinGonka API Gateway you get MiniMax M2.7, DeepSeek V4 Flash, and GLM-5.3 Flash via an OpenAI-compatible API: any code written for OpenAI works unchanged — just swap the URL, API key, and model name.

Example request to an active model:

curl https://gate.joingonka.ai/api/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMaxAI/MiniMax-M2.7",
    "messages": [{"role": "user", "content": "Explain the MoE architecture"}]
  }'

Pricing: $0.0062 per 1 million tokens — that's ~810 times cheaper than GPT-5.5 ($5.00/1M) and ~490 times cheaper than Claude Sonnet 4.6 ($3.00/1M). When you sign up, you get 3M free tokens to test with.

The Gateway works with popular development tools: Quick Start covers connecting via Python, Node.js, and curl. IDE integrations are supported too — Cursor, Continue, Cline, Aider, and Claude Code — along with AI agent frameworks: LangChain, n8n, LibreChat, Open WebUI.

To get started fast:

1. Sign up at gate.joingonka.ai (connect a wallet or create a new one)
2. Get an API key in the Dashboard
3. Replace api.openai.com with gate.joingonka.ai/api in your code
4. Specify a current network model — MiniMaxAI/MiniMax-M2.7, deepseek-ai/DeepSeek-V4-Flash-0731, or zai-org/GLM-5.3-Flash (full list at the GET /v1/models endpoint)

Enterprise-grade decentralized inference at a hobby-project price isn't going anywhere — Qwen3-235B has simply made way for the network's newer models. The same combination of price and quality now runs on MiniMax M2.7, DeepSeek V4 Flash, and GLM-5.3 Flash.