EyesInAI
Switchyard
Groundyard
Connect
Benchmarks
Models
Learn
News
Home
Switchyard
Groundyard
Connect
Which model for your task?
Leaderboard
Deep Suite
Coding Models
Bake-offs
Measured
Compare
Trends
Try Models
Catalog
Providers
By Origin
By Layer
Hardware
Open Source
License Tiers
What we’re building
Explainers
Automation & Agents
Cut Costs
Case Studies
Papers
Notes
Principles
Methodology
Jargon
Help
Contact
Latest
Governance & Access
Pricing & Billing
Data Centers & Power
Articles
Investigated & Declined
Trending models
Status
Archive
Benchmark your use case
Loading benchmark data…
EyesInAI
·
Loading live benchmark data
Home
Benchmark
Settings
Loading models…
EyesInAI
·
Loading models…
All models
fireworks
accounts/fireworks/models/kimi-k2p6
Last tested Jul 25, 2026
Overall pass
100%
Avg latency
12536 ms
Context
262k
Tools
Yes
Input $/1M
$0.95
Output $/1M
$4.00
Tests run
18
Passed
18/18
AI summary
Hardware to run it
Hosted:
$0.95 in / $4 out
per 1M
Self-host:
GPU cluster
1T total / 32B active (MoE)
FP16 ~2TB
INT4 ~542GB
Multi-node H100/H200
Trillion-param scale — cluster-only at full precision; ~542GB at INT4.
Test results
⚡
Ping
Latency & availability — single-word reply
1744 ms
🧮
Reasoning
Basic math reasoning — show work, give answer
12038 ms
{ }
JSON Output
Structured output compliance — valid JSON with required keys
2900 ms
💻
Code Gen
Python function generation with docstring
8448 ms
🚀
Throughput
Token generation speed — 500-token long-form response
7794 ms
🔍
Context Recall
Retrieval from in-context data — 20-item list Q&A
Compare with another model
See the leaderboard
8068 ms
🔧
Tool Use
Function/tool calling — get_weather invocation
2719 ms
📝
Summarization
Real-world: distill a news article into 3 bullet points
8405 ms
🏷️
Classification
Real-world: sentiment + category from customer review
8908 ms
📄
Data Extraction
Real-world: pull structured fields from an invoice
14229 ms
📋
Instruction Follow
Real-world: multi-rule compliance (5 sentences, no "the", etc.)
8703 ms
✅
Format Compliance
IFEval-style: 4 bullets, keyword inclusion/exclusion, no preamble
31708 ms
🪡
Long-Context Needle
Find a 6-digit code buried in ~3.5k tokens of filler text
2420 ms
🧠
Multi-Step Logic
BBH-lite: boolean expression + web of lies + object counting
6699 ms