Guides → Playground & Guide → Open-Weight Model Suitability - Can This Client Actually Self-Host It, and May They
Meet Dan Whitfield. vCIO at a 40-client MSP, asked about a cheap Chinese model by a client who read about it. "A client saw a headline about a model that beats the expensive ones at a fraction of the price. Everyone says just self-host it. Can they, actually, and may they?"
🔥 I need an answer today that is defensible in six months, and I do not want to say no purely because it is Chinese if the real answer is that it is fine for their marketing team.
Separate two risk classes before anything else. One set comes from where the model is hosted: jurisdiction, residency, and whether a foreign legal regime can compel disclosure. The other applies to any open model regardless of origin: prompt injection, leakage from fine-tuning on sensitive records, and undocumented provenance. Conflating them produces both errors - blocking a perfectly good European open model, and assuming self-hosting cured prompt injection. It did not.
Only the first class is fixed by self-hosting. Moving the weights inside the perimeter solves data locality and nothing else.
Then check whether self-hosting is real for this client. A mixture-of-experts model pays memory for its total parameters and compute only for the active ones, so all experts must be resident. That is what puts a 2.8 trillion parameter model beyond a single well-specified workstation regardless of how few parameters activate per token.
And check the licence. Weights can be downloadable under terms that restrict commercial use, or that attach obligations above a revenue or user threshold. Read the actual file rather than assuming the word open settles it.
Same calculator, different applications. Sizes above, workloads here. Pick the one that looks like yours.
Pre-loaded scenarios for the most common applications. Click a tab to see realistic numbers, then hit "Try this scenario" to load it into the calculator above.
Open the full calculator. Pick a model, enter your tokens, see per-call, daily, monthly, and annual cost.
🚀 Open the full calculator →Answers whether an open-weight model is genuinely deployable for a client workload. Checks whether the weights shipped, whether the licence permits commercial use, whether the hardware exists, and whether the data class allows the hosted route at all.
Gate 1, weights shipped. Not announced, not promised, present and downloadable. If not, the only available route is the hosted API and the data class decides everything.
Gate 2, licence permits this use. Commercial use allowed, and any threshold obligations checked against the client's actual scale.
Gate 3, hardware feasible. Estimated weight memory against what the client has or would rent. Fails far more often than people expect.
Gate 4, data class. Green, amber or red for the hosted route, using the same bands as jurisdiction-risk-scorer.
The output is a verdict per route, not per model: hosted API, rented GPU in a chosen region, private cloud under an existing agreement, or on-premises.
Honest limitations. Every model is wrong; some are useful. Where this one falls short:
For these, use: Run jurisdiction-risk-scorer to turn the data class into a green, amber or red verdict per endpoint,Run msp-client-ai-exposure-audit to find whether a router can already reach the endpoint you just decided against,Run local-vs-cloud-inference to compare a feasible self-host against the hosted price,Run self-host-breakeven for the volume at which owning hardware beats renting it
The gaps we just listed are real, and they are the expensive ones: your actual prompts, your switching cost, your MLOps overhead. An AICost expert spends the hour on your AI and cloud costs, not a generic playbook. You leave with a written report: the way forward, in 30 days of concrete steps.
Not sure yet? The $39 AICost Blueprint credits toward a Session, and the Session fee credits toward any plan. You never pay twice for the same ground. See all pricing →
You have seen the shape of it. Open the calculator with your model, your tokens, your volume.
🚀 Open the full calculator →Author: Subu Vdaygiri, Founder & CEO of CloudIntelligence.ai. 17 years Fortune 100 (Ingram Micro, Siemens). Wharton CTO program · Kellogg CPO program · 10× AWS+Azure certified.
Last-verified date is the most recent successful daily snapshot
(aicost_pricing_snapshots) or, when no snapshot exists yet,
the latest successful crawler run (aicost_crawler_runs).
10 of 10
vendors are currently verified. Aggregator services (TokenCost, AI Pricing Guru, etc.)
are not listed.
Derived from industry conventions, not directly published by the vendor. Typical conventions: cached input = 10% of base (90% off), Batch API = 50% of base (50% off).
| Vendor / Model | Field | Why it’s inferred |
|---|---|---|
| Anthropic — Claude Sonnet 4.6 | cachedInput |
Derived at 10% of input rate — Anthropic publishes 90% cache-hit discount on this tier. |
| Anthropic — Claude Sonnet 4.5 | cachedInput |
Derived at 10% of input rate; same 90% cache-hit convention as Sonnet 4.6. |
| Anthropic — Claude Sonnet 4.5 | batchInput |
Derived at 50% of standard input — Anthropic documents uniform 50% Batch discount. |
| Anthropic — Claude Sonnet 4.5 | batchOutput |
Derived at 50% of standard output — Anthropic documents uniform 50% Batch discount. |
| Anthropic — Claude Haiku 4.5 | cachedInput |
Derived at 10% of input rate — Anthropic 90% cache-hit discount convention. |
| OpenAI — GPT-5.4 Mini | cachedInput |
Derived at 10% of input — OpenAI documents automatic 90% discount on cache hits across GPT-5.x tier. |
| OpenAI — GPT-5.4 Nano | cachedInput |
Derived at 10% of input — OpenAI 90% cache-hit convention. |
| OpenAI — GPT-5.4 Nano | batchInput |
Derived at 50% of input — OpenAI Batch API uniform 50% discount. |
| OpenAI — GPT-5.4 Nano | batchOutput |
Derived at 50% of output — OpenAI Batch API uniform 50% discount. |
| OpenAI — GPT-5.4 Pro | cachedInput |
Derived at 10% of input — OpenAI 90% cache-hit convention. |
| OpenAI — GPT-5.4 Pro | batchInput |
Derived at 50% of input — OpenAI Batch API uniform 50% discount. |
| OpenAI — GPT-5.4 Pro | batchOutput |
Derived at 50% of output — OpenAI Batch API uniform 50% discount. |
| OpenAI — GPT-5.2 | cachedInput |
Derived at 10% of input; no residency uplift. |
| OpenAI — GPT-5.2 | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5.2 | batchOutput |
Derived at 50% of output. |
| OpenAI — GPT-5 | cachedInput |
Derived at 10% of input. |
| OpenAI — GPT-5 | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5 | batchOutput |
Derived at 50% of output. |
| OpenAI — GPT-5.5 Pro | cachedInput |
Derived at 10% of input — OpenAI does not publish a cached rate for *-pro models; using the family convention. |
| OpenAI — GPT-5.5 Pro | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5.5 Pro | batchOutput |
Derived at 50% of output. |
| OpenAI — GPT-5.2 Pro | cachedInput |
Derived at 10% of input — pro-tier convention. |
| OpenAI — GPT-5.2 Pro | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5.2 Pro | batchOutput |
Derived at 50% of output. |
| OpenAI — GPT-5.1 | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5.1 | batchOutput |
Derived at 50% of output. |
| OpenAI — GPT-5 Pro | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5 Pro | batchOutput |
Derived at 50% of output. |
| OpenAI — GPT-5 Nano | cachedInput |
Derived at 10% of input. |
| OpenAI — GPT-5 Nano | batchInput |
Derived at 50% of input. |
| OpenAI — GPT-5 Nano | batchOutput |
Derived at 50% of output. |
| Google — Gemini 3 Flash | cachedInput |
Derived at 10% of input — Google caching discount convention ~90%. |
| Google — Gemini 3.1 Flash-Lite | cachedInput |
Derived at 10% of input — Google caching convention. |
| Google — Gemini 3.1 Flash-Lite | batchInput |
Derived at 50% of input — Google Batch API uniform 50% discount. |
| Google — Gemini 3.1 Flash-Lite | batchOutput |
Derived at 50% of output — Google Batch API uniform 50% discount. |
| Google — Gemini 2.5 Pro | cachedInput |
Derived at 10% of input. |
| Google — Gemini 2.5 Flash | cachedInput |
Derived at 10% of input. |
| Google — Gemini 2.5 Flash-Lite | cachedInput |
Derived at 10% of input — Google caching convention. |
| Google — Gemini 2.5 Flash-Lite | batchInput |
Derived at 50% of input — Google Batch API uniform 50% discount. |
| Google — Gemini 2.5 Flash-Lite | batchOutput |
Derived at 50% of output — Google Batch API uniform 50% discount. |
| Google — Gemini 2.0 Flash | cachedInput |
Derived at 25% of input per Google 2.0 family caching rates. |
| Google — Gemini 2.0 Flash | batchInput |
Derived at 50% of input — Google Batch API uniform 50% discount. |
| Google — Gemini 2.0 Flash | batchOutput |
Derived at 50% of output — Google Batch API uniform 50% discount. |
| Google — Gemini 2.0 Flash-Lite | cachedInput |
Derived at 10% of input — Google caching convention. |
| Google — Gemini 2.0 Flash-Lite | batchInput |
Derived at 50% of input — Google Batch API uniform 50% discount. |
| Google — Gemini 2.0 Flash-Lite | batchOutput |
Derived at 50% of output — Google Batch API uniform 50% discount. |
| xAI — Grok 4 (legacy) | cachedInput |
Extrapolated at 25% of base. |
Pricing is cross-verified against the
LiteLLM community registry
when available. Daily snapshots are kept in aicost_pricing_snapshots;
every change is logged to aicost_price_changelog with old & new
values for full audit trail. Read the full methodology →