← Tools
Self-assessment · v0.1

Where your AI budget actually goes.

Estimate monthly inference spend across four common enterprise functions, then see how frontier labs compare to sovereign and self-hosted open weights, energy included. Refreshed August 2026 on published prices and the Artificial Analysis Intelligence Index. A directional model, useful in five minutes.

your monthly task volume
Adjust the sliders
Legal & contracts
Contract review, clause extraction, compliance memos.
800
tasks / mo
12k in · 4k out per task
Marketing & creative
Briefs, variants, multi-channel copy, campaign QA.
3,500
tasks / mo
3k in · 6k out per task
HR & compliance
Policy Q&A, screening, internal comms, training.
1,200
tasks / mo
5k in · 2k out per task
Coding & app dev
Codegen, refactors, test scaffolds, code review.
5,000
tasks / mo
8k in · 10k out per task
if you self-host, your hardware
The part rented models hide in the token price
GPU server, each
$60k
Slide to zero if the hardware is already paid for
Written down over
36 mo
1 node at your volume · 143M tokens / mo
Estimated monthly inference
Claude Opus 5
Frontier · US
$2.2k/mo
DeepSeek V4 Pro
Sovereign EU cloud
$149/mo
Qwen3.8 27B
Self-hosted · your DC
$1.7k/mo
≈ 22% cheaper, self-hosted.
Includes $1.7k/mo of hardware (1 node written down over 36 months) and EU energy. Excludes ops headcount.
Against renting the same class of open weights, self-hosting is 11.7x the cost. That comparison, not the frontier one, is what the hardware decision turns on.
Owning the box overtakes renting at 3.2B tokens a month, about 83% of what this hardware can serve. Below that you are paying for idle capacity; above it the box wins and keeps winning.
cost is half the story, scroll for performance by task.
Per-function breakdown

Cost and intelligence, side by side.

For each function, we show monthly cost per model and a directional performance score. The right answer is rarely the cheapest, it's the cheapest model that clears the quality bar for that task.

Legal & contracts

800 tasks/mo
Claude Opus 5
$128
AA 63
DeepSeek V4 Pro
$9.0
AA 52
Qwen3.8 27Bcheapest
$6.2
AA 51

Marketing & creative

3,500 tasks/mo
Claude Opus 5
$577
AA 62
DeepSeek V4 Pro
$38
AA 51
Qwen3.8 27Bcheapest
$17
AA 51

HR & compliance

1,200 tasks/mo
Claude Opus 5
$90
AA 62
DeepSeek V4 Pro
$6.2
AA 52
Qwen3.8 27Bcheapest
$4.1
AA 52

Coding & app dev

5,000 tasks/mo
Claude Opus 5
$1.4k
AA 65
DeepSeek V4 Pro
$96
AA 57
Qwen3.8 27Bcheapest
$48
AA 53
Cost is per-token, plus the box.
Self-hosting now carries amortised GPU capex and EU energy, sized to your volume. Still excludes ops headcount. The per-function bars below show marginal cost only.
Quality is the published index.
Artificial Analysis Intelligence Index, Aug 2026 (Opus 5 leads at 63). The small per-function tilt is our judgment, not theirs.
Volumes are yours to set.
Defaults reflect a mid-size enterprise. Adjust to your reality.

Sources, August 2026: frontier pricing from Anthropic's published API rates; open-weight pricing and the quality index from Artificial Analysis. The self-hosted line is modelled rather than quoted: the same Qwen3.8 27B weights rent at $0.40 in and $3.00 out per million tokens on a public host, and owning the hardware removes that margin, which is the whole comparison. The open-weight table moves every few weeks, so treat any figure here as a starting point rather than a quote.

Get the full report

Your numbers, properly modelled.

We'll send you a PDF with your scenario, refined with realistic GPU capex, energy under EU industrial tariffs, ops headcount, and the break-even point against frontier-only spend. Plus the per-function recommendation we'd give in a working session.

your scenario stays in your browser. only your contact details are sent.
Request this research

My inference cost report