LIVE
Nov 11New model: ERNIE-5.5-Flash-Preview by baidu·Nov 11New model: ERNIE-Light-DT-1.0 by baidu·Nov 9New model: Morph V3 Large (batch) by morph·Nov 9New model: Morph V3 Fast (batch) by morph·Nov 9DeepSeek V4 Pro pricing increased 27%·Nov 9Qwen3.5 397B A17B pricing increased 20%·Nov 9GLM 5.2 (batch) pricing increased 68%·Nov 8DeepSeek V4 Flash 0731 pricing dropped 26%·Nov 8DeepSeek V4.1 Flash (batch) pricing dropped 90%·Nov 8Qwen3.6 27B pricing dropped 29%·Nov 8DeepSeek V4.1 Flash 1017 pricing dropped 96%·Nov 8GLM 5.3 Flash pricing dropped 13%·Nov 8DeepSeek V4 Pro 0813 pricing dropped 5%·Nov 8Kimi K2.5 pricing increased 33%·Nov 7New model: LFM3-1.7B (free) by liquid·Nov 7New model: Qwen3.8 27B (batch) by Alibaba Qwen·Nov 7MiniMax M2.5 pricing increased 26%·Nov 7Gemma 4 26B A4B pricing dropped 12%·Nov 7Nemotron 3 Super pricing dropped 83%·Nov 7DeepSeek V4.1 Flash pricing increased 13%·Nov 7DeepSeek V3 pricing increased 24%·Nov 7MiniMax M3 (batch) pricing dropped 27%·Nov 6Qualcomm reported fiscal-year revenue of $43.1B·Nov 6New model: Gemini Omni 1.2 Flash by Google DeepMind·Nov 6New model: Gemini Omni 1.1 Flash by Google DeepMind·Nov 6New model: Auto Code by openrouter·Nov 6DeepSeek V4 Flash Vision Exp pricing increased 104%·Nov 6Beam 1.0 pricing increased 20%·Nov 6MiniMax M2.7 pricing dropped 30%·Nov 4New model: Hy4 by tencent·Nov 4New model: Grok 4.7 (batch) by xAI·Nov 4New model: Grok 4.6 (batch) by xAI·Nov 4Qwen3 Coder 30B A3B Instruct pricing increased 114%·Nov 4MiniMax M2.5 pricing dropped 21%·Nov 4DeepSeek V4 Flash 0731 pricing dropped 50%·Nov 4Granite 4.2 8B pricing increased 67%·Nov 4Qwen3.8 27B pricing increased 285%·Nov 4Kimi K2.7 Code pricing increased 13%·Nov 4Qwen3.6 27B pricing increased 41%·Nov 4gpt-oss-20b pricing increased 67%·Nov 4DeepSeek V4 Pro pricing increased 96%·Nov 4Gemma 4 26B A4B pricing dropped 7%·Nov 4Kimi K3 pricing increased 7%·Nov 4GLM 5.3 Flash pricing dropped 39%·Nov 4DeepSeek V4 Pro 0813 pricing dropped 91%·Nov 2New model: GLM 5.3 FlashX Legal by z-ai·Oct 30Mistral Large 4 pricing dropped 50%·Oct 30DeepSeek V4 Pro pricing dropped 15%·Oct 30Kimi K3 pricing dropped 25%·Oct 30Qwen3.6 27B pricing dropped 29%·Nov 11New model: ERNIE-5.5-Flash-Preview by baidu·Nov 11New model: ERNIE-Light-DT-1.0 by baidu·Nov 9New model: Morph V3 Large (batch) by morph·Nov 9New model: Morph V3 Fast (batch) by morph·Nov 9DeepSeek V4 Pro pricing increased 27%·Nov 9Qwen3.5 397B A17B pricing increased 20%·Nov 9GLM 5.2 (batch) pricing increased 68%·Nov 8DeepSeek V4 Flash 0731 pricing dropped 26%·Nov 8DeepSeek V4.1 Flash (batch) pricing dropped 90%·Nov 8Qwen3.6 27B pricing dropped 29%·Nov 8DeepSeek V4.1 Flash 1017 pricing dropped 96%·Nov 8GLM 5.3 Flash pricing dropped 13%·Nov 8DeepSeek V4 Pro 0813 pricing dropped 5%·Nov 8Kimi K2.5 pricing increased 33%·Nov 7New model: LFM3-1.7B (free) by liquid·Nov 7New model: Qwen3.8 27B (batch) by Alibaba Qwen·Nov 7MiniMax M2.5 pricing increased 26%·Nov 7Gemma 4 26B A4B pricing dropped 12%·Nov 7Nemotron 3 Super pricing dropped 83%·Nov 7DeepSeek V4.1 Flash pricing increased 13%·Nov 7DeepSeek V3 pricing increased 24%·Nov 7MiniMax M3 (batch) pricing dropped 27%·Nov 6Qualcomm reported fiscal-year revenue of $43.1B·Nov 6New model: Gemini Omni 1.2 Flash by Google DeepMind·Nov 6New model: Gemini Omni 1.1 Flash by Google DeepMind·Nov 6New model: Auto Code by openrouter·Nov 6DeepSeek V4 Flash Vision Exp pricing increased 104%·Nov 6Beam 1.0 pricing increased 20%·Nov 6MiniMax M2.7 pricing dropped 30%·Nov 4New model: Hy4 by tencent·Nov 4New model: Grok 4.7 (batch) by xAI·Nov 4New model: Grok 4.6 (batch) by xAI·Nov 4Qwen3 Coder 30B A3B Instruct pricing increased 114%·Nov 4MiniMax M2.5 pricing dropped 21%·Nov 4DeepSeek V4 Flash 0731 pricing dropped 50%·Nov 4Granite 4.2 8B pricing increased 67%·Nov 4Qwen3.8 27B pricing increased 285%·Nov 4Kimi K2.7 Code pricing increased 13%·Nov 4Qwen3.6 27B pricing increased 41%·Nov 4gpt-oss-20b pricing increased 67%·Nov 4DeepSeek V4 Pro pricing increased 96%·Nov 4Gemma 4 26B A4B pricing dropped 7%·Nov 4Kimi K3 pricing increased 7%·Nov 4GLM 5.3 Flash pricing dropped 39%·Nov 4DeepSeek V4 Pro 0813 pricing dropped 91%·Nov 2New model: GLM 5.3 FlashX Legal by z-ai·Oct 30Mistral Large 4 pricing dropped 50%·Oct 30DeepSeek V4 Pro pricing dropped 15%·Oct 30Kimi K3 pricing dropped 25%·Oct 30Qwen3.6 27B pricing dropped 29%·
The AI Economy, Tracked.
Today in AI · GLM 5.2 (batch) pricing increased 68%
Pulse40·frothy
Bubble255%·frothy
Open Source60.4%
Frontier Race
Claude Fable 5.2
#1
95.4
$8.00/M
DeepSeek V3.2 Speciale
#2
95.2
$0.40/M
GPT-6 Astra
#3
93.4
$10.00/M
Step 3.5 Flash
#4
89.5
$0.10/M
GPT-5 Chat
#5
89.0
$1.25/M
Claude Opus 5.5
#6
88.3
$4.00/M
GPT-5.5 Pro
#7
87.6
$30.00/M
GPT-6.1 Sol
#8
87.4
$2.00/M
GPT-5.1-Codex-Max
#9
85.6
$1.25/M
Gemini 4 Argon
#10
84.5
$2.00/M
Claude Opus 4.6 (Fast)
#11
83.3
$30.00/M
Claude Opus 5
#12
83.2
$5.00/M
GPT-6.1 Terra
#13
82.8
$1.00/M
Bar width · avg benchmark score · Color · category
Mindshare Arena
Tile size · share of voice · green · rising this week · red · falling
The Pulse
Gecko Pulse
40
frothy
7d · +6 pts
healthyfrothyoverheatedbubble
Bubble Index · components
Valuation Premiumhealthy62
Funding Accelerationmoderate138
Concentration Riskhealthy23
Revenue Qualityhealthy7
Capex · Revenue Gaphealthy26
Largest driver · Funding Acceleration · 138 of 255 pts · company data as of 2026-11-10
AI Bubble Index255%
HealthyFrothyOverheatedBubble
Gecko Tests · our own data
From the Lab
Day 1Claude Fable 5.2 launch report card · BenchGecko's own tests on day oneFinding · 2026-11-10Day 1Gemini 4 Carbon launch report card · BenchGecko's own tests on day oneFinding · 2026-11-10-90%DeepSeek V4.1 Flash (batch) price cut 90%Finding · 2026-11-09All findingsTop on our tests1. Qwen3.8 Max Prime 832. Qwen3.8 27B 783. Qwen3.8 2.4T A95B 75
Cross-layer insights
Latest Benchmark Results
All →Today in AI
All →Nov 11
New model: ERNIE-5.5-Flash-Preview by baiduNov 11New model: ERNIE-Light-DT-1.0 by baiduNov 9
New model: Morph V3 Large (batch) by morphNov 9New model: Morph V3 Fast (batch) by morphNov 9
DeepSeek V4 Pro pricing increased 27%Nov 9
Qwen3.5 397B A17B pricing increased 20%Nov 9
GLM 5.2 (batch) pricing increased 68%Nov 8DeepSeek V4 Flash 0731 pricing dropped 26%Nov 8DeepSeek V4.1 Flash (batch) pricing dropped 90%Nov 8Qwen3.6 27B pricing dropped 29%Nov 8DeepSeek V4.1 Flash 1017 pricing dropped 96%Nov 8GLM 5.3 Flash pricing dropped 13%
More Indicators
Gecko Pulse
40+6 7d
frothy · bubble pressure 17, attention 59.
AI Bubble Index
255%-15 7d
frothy · company data as of 2026-11-10.
Mindshare Pulse
59+36 7d
7,153 tracked HN mentions this week vs 6,079 average.
AICDI · Compute
59-2 7d
Growing pressure across the AI infrastructure stack
FCI · Foundries
830 7d
Foundry concentration index.
MDI · Memory
570 7d
Memory pressure index.
Capex / Revenue
1.3×
Tracked capex vs AI revenue · company data as of 2026-11-10.
The Matrix · Top 20 Models
| # | Benchmarks | ||||
|---|---|---|---|---|---|
| 1 | 95.4 | $8.00 | 1000K | 10 | |
| 2 | 95.2 | $0.40 | 164K | 9 | |
| 3 | 93.4 | $10.00 | 1050K | 34 | |
| 4 | 89.5 | $0.10 | 262K | 10 | |
| 5 | 89.0 | $1.25 | 128K | 7 | |
| 6 | 88.3 | $4.00 | 1000K | 28 | |
| 7 | 87.6 | $30.00 | 1050K | 14 | |
| 8 | 87.4 | $2.00 | 1050K | 26 | |
| 9 | HA Qwen2.5 72B Instruct Abliterated | 87.2 | n/a | n/a | 6 |
| 10 | 85.9 | n/a | n/a | 7 | |
| 11 | 85.6 | $1.25 | 400K | 8 | |
| 12 | 84.5 | $2.00 | 1049K | 26 | |
| 13 | 83.3 | $30.00 | 1000K | 12 | |
| 14 | 83.2 | $5.00 | 1000K | 25 | |
| 15 | 82.8 | $1.00 | 1050K | 11 | |
| 16 | 82.4 | $2.00 | 1000K | 26 | |
| 17 | U PaLM 2-L | 81.8 | n/a | n/a | 4 |
| 18 | 81.7 | $0.09 | 262K | 11 | |
| 19 | 81.3 | $10.00 | 1000K | 30 | |
| 20 | 80.7 | $1.75 | 400K | 9 |
The Stack · 11 Data Layers
Models
Benchmark, price, context · every frontier model
1310 tracked
Providers
Availability matrix across every AI inference provider
286 tracked
Agents
SWE-bench · MLE-bench · OSWorld and more
165 tracked
Benchmarks
Score evolution across every major benchmark
160 tracked
Research
Research library · methods · notes · source maps
Hub
Economy
Bubble Index · valuations · funding · market cap
71 companies
Mindshare
Kaito-style attention tracking across all signals
96 entities
Infra
Chips · foundries · memory · systems · energy
6 indices
Pricing
Arbitrage finder · substitute finder · free tier tracker
Live
Learn
251 terms · glossary · learning paths
251 terms
Compare
Universal compare · models · agents · providers
15+ routes
Learn AI
All paths →Path · 7 terms
The AI Bubble Explained
Seven terms that decode whether AI is overpriced, fairly priced, or criminally underpriced. Read in order.
Path · 6 terms
Pick an AI Model
Six terms to go from "I need an AI" to "here is the cheapest model that meets my spec."
Path · 7 terms
From Sand to Model
The AI supply chain in 7 terms · foundry, memory, chip, system, training, inference, pricing.
Featured Indicators
Gecko Pulse
AI ecosystem health · 0-100 composite
See full chart →
AI Bubble Index
Current reading vs dotcom peak
See full chart →
Mindshare Arena
Share of voice across models
See full chart →
Frontier Race
Top models by avg benchmark score
See full chart →
HBM Bandwidth
Memory supply tension index
See full chart →
Parabolic Capex
Hyperscaler AI capex trajectory
See full chart →
Methodology · FAQ · Freshness
Full methodology →How often is BenchGecko data updated?›
Models and prices refresh daily from OpenRouter, benchmark scores daily from public leaderboards (Epoch AI, Artificial Analysis, LMArena, LiveBench and others), and Mindshare daily from Hacker News, Wikipedia and GitHub. Each page shows the date its data was last updated.
What is the Gecko Pulse?›
A 0-100 heat score for the AI market: 40% Bubble Index pressure, 20% model release velocity, 20% share of premium-priced models, 20% Mindshare attention. Higher means hotter. Daily values are recorded so changes are real, not estimated.
How are benchmark scores normalized?›
Every benchmark is min-max normalized across the full set of scored models. Rankings average normalized scores across ≥3 benchmarks per model to avoid over-weighting any single test.
Where does pricing data come from?›
OpenRouter list prices, snapshotted daily. Price-change events are only published when a change holds for a full day, so routing flicker is filtered out.
Can I cite BenchGecko data?›
Yes. Every page ships a Share & Cite bar with APA, MLA, BibTeX, Chicago, and plain-text formats. Attribution is required on the free API tier and encouraged everywhere.
Sources ·OpenRouterEpoch AISWE-benchMCP RegistryChatbot ArenaHuggingFaceLiveBenchArtificial AnalysisSEALAider
Updated 2h ago · 10+ authoritative sources · zero editorial content·Learn · Glossary·Research·Developers