v0.1 · catalog 2026.11.19 · MIT

Stop overpaying for simple steps.

Your agent uses one model for everything. ModelPick picks the model, shows the real cost, and gives you the key — for coding, image, video, 3D, and audio.

$ npx -y github:Leizhidong-creator/modelpick pick "3D sword"

No npm auth: npm install modelpick --registry=https://leizhidong-creator.github.io/modelpick/npm/

zero dependencies no key needed 40 models · 20 nodes
modelpick — bash
modelpick pick "3D sword for a Unity game"
3D asset generation · high confidence Tripo H3.1 · Tripo quality 94 score 80.6/100 cost $0.21 / usable asset $0.20 raw × 1.07 attempts #2 Hunyuan3D 3.1 Pro $0.59 / asset #3 Meshy 7.1 $1.34 / asset saves 6.2× vs Meshy 7.1 connect: export TRIPO_API_KEY=...
40models · 18 providers
20task nodes · 5 modalities
0runtime dependencies
1.1msp95 decision latency
Intro video

Not the cheapest. The best value.

The problem, the 3D decision, the five-model video plan, and the evidence — with English voiceover. Captions in the repo.

English voiceover · English + 中文 captions burned in · download the bilingual SRT · script and storyboard

The cheapest token is not the cheapest task

Cheap tokens, expensive work.

A model that is 10× cheaper per token but needs 1.8× the attempts and 2× the output is not 10× cheaper. ModelPick computes the denominator that matters.

01 / COST

Cost per solved task

Price × expected attempts, cache-aware. A 48× token gap becomes a 3% cost line — and the cheap model is 27 quality points worse.

Fable 5.2 $0.63 · Haiku 5.5 $0.013
DeepSeek V4.1 Flash $0.020
02 / CAPABILITY

Modality is a filter

Text-in-image, rigging, 30s takes, lip sync, 1M context, open weights. The wrong model is rejected before it gets a score.

"headline text" → minTextInImage 90
"rigged 3D model" → requiresRigging
03 / INTEGRATION

One call inside the agent

A zero-dependency CLI, a stdio MCP server, and a portable skill. No gateway, no proxy, no key storage.

tools: pick_model · plan_workflow
surfaces: Claude · Codex · Cursor · DSH
modelpick plan "30s product launch video"
Product launch video · $0.77 estimated per run
Script → keyframes → clips → voiceover → music
1 Writing & copy Gemini 4 Hydrogen $0.03 q 81
2 Image generation Nano Banana 2.1 $0.04 q 94
3 Video generation Veo 3.1 Fast $0.60 q 86
4 Text to speech Voxtral TTS 3 $0.02 q 88
5 Music generation Lyria 3.5 $0.09 q 90
Total $0.77 per run · retries included
connect: GEMINI_API_KEY · FAL_KEY · MISTRAL_API_KEY
A model per node, not one model for everything

The launch video plan

Script → keyframes → clips → voiceover → music. Each step gets a different model, a cost estimate, and a setup block. The total is priced before you render a frame.

  • Hard constraints first: duration, audio, references, resolution
  • Alternatives with the quality and cost trade-off spelled out
  • Retries included; provider price and read date attached
  • Works offline; no key required to decide
$ modelpick plan "30s product launch video"
Coverage

Every node in the agent workflow.

Twenty task nodes, from a 400K-token codebase to a rigged game asset, with the two or three models that actually win each one.

⌘

Text

coding · review · refactor · debug · plan · long-context · research · writing · translation · extraction · classification · data · computer-use

▧

Image

generation · editing · text-in-image · 4K · character consistency

▶

Video

image-to-video · 30s takes · dialogue · native audio · 4K

◇

3D

text-to-3D · image-to-3D · PBR · rigging · quad mesh

◉

Audio

TTS · STT · voice cloning · music · timestamps

$ modelpick models --modality 3d
Install

Set up your agent.

Pick an agent. Copy the command. Restart it. The config below is what gets written — no hidden step.

Claude Code

Project MCP config, skill, and a CLAUDE.md block.

.mcp.jsonproject

            
1
Run this
2
Restart the agent

3
Ask itpick a model for a 3D sword for Unity
Tested, not guessed

Every pick has a test.

115 development runs and 35 held-out runs. Every pick, constraint, and price perturbation is in the repo.

ModelPick benchmark results
MetricDevHeld-outTarget
Top-1 acceptable100%100%≥75%
Top-3 acceptable100%100%≥92%
Hard constraints100%100%100%
Price stability100%100%≥80%
p95 latency1.1ms1.1ms≤300ms

Decision benchmark, not a model-output benchmark. Quality labels are expert-authored; media scores are priors. Full methodology and known gaps in the repo.

Built to change

Change the data, not the engine.

Prices move weekly. New providers appear monthly. The scoring engine, retry curve, and presets are JSON you can edit — with validation and a benchmark to catch regressions.

~/.modelpick/overrides.jsonprivate models
{
  "models": [{
    "slug": "my-private-model",
    "pricing": { "unit": "mtok", "input": 1, "output": 2 },
    "quality": { "overall": 88 }
  }],
  "policy": { "presets": { "cost": { "cost": 0.65 } } }
}
catalog/four files, one engine
models.json   prices · IDs · quality · sources
tasks.json    20 nodes · weights · floors
policy.json   retry curve · presets · anchors
plans.json    multi-step workflows

$ npm run validate
$ node bench/run.js
Free, MIT, no telemetry

Use it from your agent.

One command. No signup. No key needed to decide. If it saves you money, star the repo — that is the whole distribution plan.

$ npx -y github:Leizhidong-creator/modelpick pick "refactor this repo, cheapest that works"