StackSays
Menu

AI Model (LLM API) · decision tool

Which AI model should I use?

Set your monthly token volume and budget — get a ranked model verdict with the real API bill. GPT-5, Claude, Gemini, DeepSeek, priced from live data.

Prices verified 2026-07-13 against official pricing pages · fit weights are our estimates (see methodology) · share the URL to share your configuration

As of Jul 2026, with 50M input tokens/month, 10 millions output tokens / month, a $200 budget for this slot, DeepSeek V4 Pro ranks first for these inputs at 93% fit and a StackSays-estimated $30.45/month (OpenAI GPT-5 Mini: 90% fit, $32.50/month). Fit weights are our estimates; prices trace to official pages. Change any input below and the verdict recomputes.

Your situation

The verdict updates as you move these.

Narrow pick — the runner-up is close

DeepSeek V4 Profit 93% · est $30.45/mo

Open-weights workhorse at a fraction of closed-model prices

vs OpenAI GPT-5 Mini: fit 93% against 90%, est $30.45 against $32.50 per month

DeepSeek V4 Profit 93%
$30.45/mo
OpenAI GPT-5 Minifit 90%
$32.50/mo
OpenAI GPT-5fit 88%
$162.50/mo
Google Gemini 2.5 Profit 88%
$162.50/mo
DeepSeek V4 Flashfit 84%
$5.39/mo
OpenAI GPT-5 Nanofit 81%
$6.50/mo
Google Gemini 2.5 Flashfit 81%
$40.00/mo
Anthropic Claude Haiku 4.5fit 81%
$100.00/mo
Anthropic Claude Sonnet 5fit 76%
$200.00/mo
Anthropic Claude Opus 4.8fit 43%
$500.00/mo

The decision, in plain words

For a solo builder at 50M input tokens/month and a $200.00/mo budget for the AI model slot, DeepSeek V4 Pro leads at 93% fit and $30.45/mo. The two rules doing the work: budget fit (~$30.45/mo is 15% of your $200/mo budget); ecosystem maturity (established — our estimate from ecosystem age, docs and hiring pool).

OpenAI GPT-5 Mini runs close (90% fit, $32.50/mo). It takes the lead if ~$32.5/mo is 16% of your $200/mo budget matters more to you than the current weights assume.

Generated from the calculation above — every number traces to the sourced data; nothing here is written by an AI making things up.

Why DeepSeek V4 Pro scores 93%

Budget fit (weight 30, our estimate) — ~$30.45/mo is 15% of your $200/mo budget+30
Exit cost (weight 4, our estimate) — Open weights — self-host or switch inference providers freely+4
Ecosystem maturity (weight 8, our estimate) — established — our estimate from ecosystem age, docs and hiring pool+8
Capability tier (weight 14, our estimate) — workhorse — our estimate from vendor positioning, not a benchmark; budget models degrade first on long agentic chains+6.3
Total (lines sum exactly)48.3
Possible range −56…+56 (sum of active rule weights) · fit % = (total + max) / (2 × max)(48.3 + 56) / 112 = 93%

How the $30.45/mo estimate is calculated

input tokens overage — 50M over 0M included × $0.435/M$21.75
output tokens overage — 10M over 0M included × $0.87/M$8.70
Note — Prompt caching available (cached reads $0.0036/M) — not modeled here$0
Monthly total (lines sum exactly)$30.45

What could change this verdict

Cost risk

Estimated now (StackSays estimate, not your final bill)$30.45/mo · API pay-as-you-go
Free-tier statusNo free tier — paid from the first unit
Biggest unpriced/uncertain variablePrompt caching available (cached reads $0.0036/M) — not modeled here

Official prices verified 2026-07-13; calculation is deterministic code over those prices.

At a glance

ToolFree tierPaid entryVerified
Anthropic Claude Opus 4.8No free tierPay-as-you-go — Prompt caching available (cached reads $0.5/M) — not modeled here2026-07-13
Anthropic Claude Sonnet 5No free tierPay-as-you-go — Prompt caching available (cached reads $0.2/M) — not modeled here2026-07-13
Anthropic Claude Haiku 4.5No free tierPay-as-you-go — Prompt caching available (cached reads $0.1/M) — not modeled here2026-07-13
OpenAI GPT-5No free tierPay-as-you-go — Prompt caching available (cached reads $0.125/M) — not modeled here2026-07-13
OpenAI GPT-5 MiniNo free tierPay-as-you-go — Prompt caching available (cached reads $0.025/M) — not modeled here2026-07-13
OpenAI GPT-5 NanoNo free tierPay-as-you-go — Prompt caching available (cached reads $0.01/M) — not modeled here2026-07-13
Google Gemini 2.5 ProNo free tierPay-as-you-go — Prompt caching available (cached reads $0.125/M) — not modeled here2026-07-13
Google Gemini 2.5 FlashNo free tierPay-as-you-go — Prompt caching available (cached reads $0.03/M) — not modeled here2026-07-13
DeepSeek V4 ProNo free tierPay-as-you-go — Prompt caching available (cached reads $0.0036/M) — not modeled here2026-07-13
DeepSeek V4 FlashNo free tierPay-as-you-go — Prompt caching available (cached reads $0.0154/M) — not modeled here2026-07-13
Where did you land?anonymous, one click

Data sources

ToolSourceVerified
Anthropic Claude Opus 4.8OpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
Anthropic Claude Sonnet 5OpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
Anthropic Claude Haiku 4.5OpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
OpenAI GPT-5OpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
OpenAI GPT-5 MiniOpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
OpenAI GPT-5 NanoOpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
Google Gemini 2.5 ProOpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
Google Gemini 2.5 FlashOpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
DeepSeek V4 ProOpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13
DeepSeek V4 FlashOpenRouter live pricing (CalcSays daily sync, fetched 2026-07-13)2026-07-13

A monitor re-checks these pages for changes; when a vendor moves a price, the date updates and every verdict recomputes.

People also ask

How much does the Claude API cost?

From live pricing (synced 2026-07-13): Claude Sonnet 5 runs $2 in / $10 out per million tokens, Haiku 4.5 $1/$5, Opus 4.8 $5/$25. A feature doing 50M input + 10M output tokens a month costs $200 on Sonnet 5 — the calculator above prices your exact volume.

How much does the OpenAI API cost?

From live pricing (synced 2026-07-13): GPT-5 runs $1.25 in / $10 out per million tokens, GPT-5 mini $0.25/$2, GPT-5 nano $0.05/$0.40. Batch API halves these for non-realtime jobs. Set your volume above for the real monthly number.

What is the cheapest LLM API?

At publish time the floor among curated models is DeepSeek v4 Flash at $0.077/$0.154 per million tokens (open weights) — roughly 16–65× cheaper than flagship closed models. Whether it's cheap enough in QUALITY for your task is the real question; budget tiers degrade first on long agentic chains.

Keep deciding