Inadviser

Price against performance, one job at a time

Leaderboards that rank what has been measured, show what it costs, and say so when the evidence is too thin to choose.

Public methodEvery source and its dateNo ads, no paid placement

AI tools

  • AI agents

    Shortlist

    57 options · 57 measured

    3 options cannot be separated.

    Shortlist, not ranked apart

    1. Claude Opus 5.5 (Max)90.1
    2. Gemini 4 Carbon (High)83.1
    3. Claude Sonnet 5.5 (Max)79.2
    What would change it

    Scores come from a CC BY dataset; cost is missing for some models. A confirmed price for each model narrows the shortlist.

  • Chat models

    Shortlist

    21 options · 21 measured

    3 options cannot be separated.

    Shortlist, not ranked apart

    1. Gemini 4 Carbon (High)80.0
    2. Gemini 4 Argon (High)76.2
    3. Claude Opus 5.5 (High)71.9
    What would change it

    Quality comes from one rating source, so the verdict stays a shortlist. Confirmed prices for the unpriced models, or a second quality source, would narrow it.

  • Image generation

    Shortlist

    19 options · 19 measured

    2 options cannot be separated.

    Shortlist, not ranked apart

    1. GPT-Image-2.5 Sunburst85.4
    2. GPT-Image-2.6 Sunburst (Medium)81.8
    What would change it

    Quality comes from one rating source, so the verdict stays a shortlist. Confirmed prices for the unpriced models, or a second quality source, would narrow it.

  • Video generation

    Shortlist

    16 options · 16 measured

    4 options cannot be separated.

    Shortlist, not ranked apart

    1. gemini-omni-1.1-flash91.6
    2. gemini-omni-flash90.7
    3. flux-3-video85.8

    and 1 more on the shortlist

    What would change it

    Quality comes from one rating source, so the verdict stays a shortlist. Confirmed prices for the unpriced models, or a second quality source, would narrow it.

  • AI search

    Shortlist

    19 options · 19 measured

    6 options cannot be separated.

    Shortlist, not ranked apart

    1. gpt-5.6-sol-xhigh84.5
    2. qwen3.8-max84.4
    3. claude-opus-4-6-search80.5

    and 3 more on the shortlist

    What would change it

    Quality comes from one rating source, so the verdict stays a shortlist. Confirmed prices for the unpriced models, or a second quality source, would narrow it.

  • Voiceover

    Insufficient evidence

    5 boards, one per job · Quality not measured. Cost only.

    What would change it

    Measured quality arrives with the first own listening tests; until then these boards show cost for the job.

  • Speech to text

    Not open yet

    Nothing published yet.

    What would change it

    Opens after the voiceover boards are stable. It would use an open test set (CC BY) and our own runs.

  • Coding agents

    Not open yet

    Nothing published yet.

    What would change it

    Needs a decision on which cost and task data may be published.

Beyond AI

  • Electric and hybrid cars

    Insufficient evidence

    30 options · 0 measured

    Quality not measured. Cost only.

    Guide: which tests to trust
    What would change it

    The board covers energy cost in Spain only, with no quality score. Reusable prices or real-world consumption data would widen it.

  • Smartphones

    Guide

    No board yet. The guide compares 5 test labs: what each measures and who funds it.

    What would change it

    The labs listed do not publish data we may reuse. A board needs a data route: a licence we can pay for, or our own tests.

  • Smartwatches

    Guide

    No board yet. The guide compares 5 test labs: what each measures and who funds it.

    What would change it

    The labs listed do not publish data we may reuse. A board needs a data route: a licence we can pay for, or our own tests.

  • Headphones

    Guide

    No board yet. The guide compares 6 test labs: what each measures and who funds it.

    What would change it

    The labs listed do not publish data we may reuse. A board needs a data route: a licence we can pay for, or our own tests.

  • Running shoes

    Guide

    No board yet. The guide compares 5 test labs: what each measures and who funds it.

    What would change it

    A board needs a licence for lab data or our own teardown tests, after citation is shown on an earlier category. Any cost needs the owner's approval first.

  • Bank offers

    Elsewhere

    Open BankJar BankJar has the same owner as Inadviser.

    What would change it

    Bank offers are not hosted on this site. They live on BankJar.

Each card is built from the boards and guides that exist. A guide lists existing tests; it is not a ranking. Also: which sites AI assistants cite today.

Data credits

Arena data: Arena (lmarena-ai), dataset lmarena-ai/leaderboard-dataset, https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset, licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Changes by Inadviser: the Arena scores (the Agent Arena IPS score and the text, text-to-image, text-to-video and search ratings) are rescaled to a 0-100 utility against fixed anchors set by Inadviser; the confidence interval and the observation or vote count are carried over; only a selection of models is shown on each board; where a board shows Arena's cost per task, it is shown unchanged. The dataset maintainers' method applies (https://arena.ai/blog/agent-arena-methodology/ for Agent Arena). Arena does not endorse Inadviser.

Energy cost per 100 km is computed by Inadviser from three inputs: consumption from the EEA, fuel prices from MITECO, and editorial electricity prices dated 2026-11-05 (EUR 0.12 and 0.20 per kWh at home, EUR 0.35 and 0.55 in public) that come from no dataset. The credits for the two datasets follow.