Skip to content T/TTokens or Towers

Compare

LeaderboardEvery tracked model, ranked and priced side by side BenchmarksWhat each test measures and who leads it PricesAPI, media, and hardware rates with a cost calculator

Plan

Stack plannerSix questions to a priced local and hosted stack ShortlistOur picks for subscriptions, APIs, and hardware EnterpriseSeat plans against self-hosting for a whole team

Build

HardwareGPUs and workstations, and which models they fit HostsWhere to rent open-weight inference, and for how much HarnessesCoding agents and the models they drive

Lab

FindingsResults from our own evaluations ScanAudit a repo’s AI usage in the browser MethodWhere the numbers come from and how we check them

Data checked September 25, 2026

Data checked September 25, 2026

T/TTokens or Towers

An independent guide to AI models, what they cost, and the hardware to run them. No account, no API key, no sponsors.

Compare

Leaderboard
Benchmarks
Prices

Plan

Stack planner
Shortlist
Enterprise

Build

Hardware
Hosts
Harnesses

Lab

Findings
Scan
Method

Data checked September 25, 2026 · Every number links to its sourceRSS · Source on GitHub

Leaderboard/Microsoft

Microsoft · Phi-4

Phi-4 Multimodal Instruct

Open weightsGenerally availableMIT

Open 5.6B model for text, image and speech understanding.

AA Intelligence: 5.8#133 of 142
Price: Self-host
Output speed: 17tok/s
Context window: 128K tokens

Composite indexes

Artificial Analysis Intelligence Index#133 of 142
6independentIntelligence Index v4.3; non-reasoning; estimatedArtificial Analysis: Phi-4 Multimodal Instruct ↗
Leader: Claude Opus 5.5 at 58

Coding

LiveCodeBench#54 of 56
13.1%independentnon-reasoningArtificial Analysis: Phi-4 Multimodal Instruct ↗
Leader: Fugu at 92.9%

Reasoning and knowledge

GPQA Diamond#140 of 142
31.5%independentnon-reasoningArtificial Analysis: Phi-4 Multimodal Instruct ↗
Leader: GPT-6 Astra at 96.1%
Humanity’s Last Exam#121 of 147
5%independentno tools; non-reasoningArtificial Analysis: Phi-4 Multimodal Instruct ↗
Leader: Claude Opus 5.5 at 61.4%

Vision and long context

MMMU-Pro#72 of 72
14.5%independentnon-reasoningArtificial Analysis: Phi-4 Multimodal Instruct ↗
Leader: Claude Opus 5.5 at 87.7%

Specs

Released: February 26, 2025
Parameters: 5.6B
Size class: Under 10B
Input: text, image, audio
Output: text
Context: 128K tokens
Max output: Not published
First token: 0.83s

Speed: Artificial Analysis leaderboard (Phi-4 Multimodal Instruct, median across providers or first-party as shown) ↗

Will it fit your budget?

The planner prices this model against local hardware and other APIs for your workload.

Plan a stack →

Compare with

Jamba Reasoning 3B5.7
Phi-45.9
Amazon Nova Micro5.9
LFM2 24B A2B5.9