Featherless Launches Fixed-Fee Private Cloud for GLM 5.2 on AMD Hardware

Featherless has launched a dedicated private cloud deployment of Z.ai's open-weight GLM 5.2 model at a flat monthly fee of $7,500, with no per-token charges. The service is optimized to run natively on AMD Instinct MI325X GPUs, and the company claims it can cut inference costs by 94% compared to closed-source rivals like GPT-5.5 and Claude Opus 4.8 . Here is everything you need to know about the service, the model, and how it stacks up against proprietary alternatives.

What is the Featherless GLM 5.2 Private Cloud?

Featherless launched a dedicated private cloud deployment of Z.ai's open-weight GLM 5.2 model, optimized to run natively on AMD Instinct MI325X GPUs at a flat monthly fee . Featherless claims it is the only platform that has optimized GLM 5.2 to run natively on AMD hardware in a private cloud environment . This approach lets customers avoid token-based billing and instead pay a fixed annual cost of USD $90,000 for a fully utilized development team .

Pricing: $7,500/Month Flat Fee with Unlimited Tokens

• Flat monthly fee: $7,500/month for a dedicated private cloud instance .
• Unlimited tokens — no per-token API charges . This is the key differentiator from pay-as-you-go APIs.
• For context, Z.ai's public API charges $1.40 per million input tokens and $4.40 per million output tokens . A heavy agentic workload costing $125,000+/month on proprietary APIs would cost roughly $7,500/month on the fixed-fee plan — a 94% reduction .

How the math works: GLM 5.2 on the public API costs about 1/5 to 1/6 the per-token price of GPT-5.5 or Claude Opus 4.8 . On pure output, GLM-5.2 at $4.40 per million tokens against GPT-5.5's $30 is about 1/6.8 the cost . Cached input drops further to $0.26 per million tokens .

Hardware: 4× AMD Instinct MI325X GPUs

Each private cloud instance runs on 4 × AMD Instinct MI325X GPUs . The MI325X features:

• 256 GB of HBM3e memory and 6 TB/s memory bandwidth per accelerator .
• CDNA3 architecture with 19,456 stream processors and 1216 matrix cores .
• FP8 performance of 2.61 PFLOPs .

This is a notably different hardware strategy from the NVIDIA H100/H200 deployments common in the industry, and Featherless frames it as a way to avoid NVIDIA supply constraints .

GLM 5.2 Model Specifications

Specification	Details
Architecture	744B total parameters, Mixture-of-Experts (MoE) with ~40B active per token
Context window	Up to 256K tokens on public cloud; up to 1 million tokens on private cloud deployments
Quantization	FP8 by default (~750 GB VRAM for weights)
Open weights	Yes — Z.ai released GLM 5.2 as open-weights under MIT license
Training focus	Coding-first; specialized for software engineering tasks

At FP8 quantization, the model's weights require roughly 750 GB of VRAM, which is comfortably accommodated by the 1 TB total VRAM across 4× MI325X GPUs (4 × 256 GB), with headroom for KV cache at extended context lengths .

Key Benchmarks: GLM 5.2 vs. GPT-5.5 and Claude Opus 4.8

SWE-bench Pro (real-world software engineering):

• GLM 5.2: 62.1% vs. GPT-5.5: 58.6% — beats GPT-5.5 on coding benchmarks .
• Trails Claude Opus 4.8 (69.2%) on this metric .

Terminal-Bench 2.1 (agentic coding tasks):

• GLM 5.2: 82.7 (best reported harness) vs. Claude Opus 4.8: 78.9 — pulls ahead on terminal-based agentic tasks .

GLM 5.2 ranks #2 on the Arena WebDev coding leaderboard according to Featherless .

Cost Comparison: GLM 5.2 vs. Proprietary APIs

The cost advantage is where GLM 5.2 really stands out. At Z.ai's official API rates:

Model	Input (per 1M tokens)	Output (per 1M tokens)
GLM 5.2	$1.40	$4.40
GPT-5.5	~$5.00	~$30.00
Claude Opus 4.8	~$8.00	~$40.00

Sources:

On a realistic 3:1 output-to-input workload blend, GLM-5.2 lands near $3.65 per million tokens against GPT-5.5's roughly $23.75 — a ratio of about 1/6.5 . Independent trackers list a lower median across providers serving the open weights (closer to ~$0.55 input and ~$1.85 output) .

Competitive Positioning vs. Proprietary AI

• Coding performance crown: GLM 5.2 beats GPT-5.5 on SWE-bench Pro while being dramatically cheaper on a per-token basis, positioning it as a high-performance, low-cost coding alternative .
• Agentic workload sweet spot: The fixed-fee private cloud is designed for sustained agentic usage (high-volume, continuous inference), where per-token API costs from proprietary vendors spiral .
• Open-weights + AMD advantage: Because the model is open and runs on AMD hardware (rather than NVIDIA H100/H200), Featherless can offer a private cloud at a fraction of the cost of equivalent NVIDIA-based deployments .
• Context window leadership: At 1M tokens on private cloud, it competes with or exceeds the context windows of GPT-5.5 and Claude Opus 4.8 on dedicated deployments .
• Caveat: On pure benchmark scores (SWE-bench Pro), Claude Opus 4.8 still leads (69.2 vs. 62.1), so GLM 5.2 is not the outright leader on every metric — but its cost advantage is massive .

Who Should Consider This Service?

The Featherless private cloud for GLM 5.2 is most compelling for:

• Teams running high-volume agentic workloads where per-token API costs quickly exceed $10,000–$50,000/month.
• Organizations with sustained coding and software engineering inference needs who want GPT-5.5-beating performance at a fraction of the cost.
• Companies looking to avoid NVIDIA hardware supply constraints by deploying on AMD infrastructure.
• Engineering teams that want unlimited token usage without worrying about metered billing.

Featherless also offers lower-tier flat-rate plans starting at $25/month for serverless access to smaller models, but the $7,500/month dedicated node is explicitly for teams needing sustained, high-volume inference on the full GLM 5.2 model .

Bottom Line

Featherless's new service offers a $7,500/month flat-fee private cloud for GLM 5.2 on 4× AMD MI325X GPUs, claiming 94% cost savings over proprietary APIs for high-volume agentic usage. The 744B MoE model beats GPT-5.5 on SWE-bench Pro at roughly one-sixth the per-token cost, making it a compelling open-weights alternative for organizations with heavy coding and agentic inference workloads. While Claude Opus 4.8 still leads on some benchmarks, the cost advantage of the GLM 5.2 + Featherless combination is large enough to make it a serious option for budget-conscious teams that don't want to compromise on performance.