This week in AI · 2026-10-12 to 2026-10-18
Four new models were listed, led by Gemini 4 Argon and Step 5 Preview. BenchGecko recorded 26 price changes, 9 of them drops, and added benchmark results for 29 models.
Week at a glance
- New models: 4
- Price changes: 26, of which 9 were price drops
- Models with benchmark results added to BenchGecko: 29
- Company updates: 3
New models
All four were dated 2026-10-15. Prices are per M tokens.
- Step 5 Preview · stepfun · input 1 · output 2.7 · context 1048576 tokens
- Gemini 4 Argon (batch) · Google DeepMind · input 1 · output 5 · context 1048576 tokens
- Gemini 4 Argon · Google DeepMind · input 2 · output 10 · context 1048576 tokens
- LFM2.5-24B-A2B · liquid · input 0.05 · output 0.7 · context 128000 tokens
Price changes
The largest changes by percentage were all input price increases (per M tokens):
- GLM 5.2 · 0.02 to 0.62 · +3000% · 2026-10-16
- DeepSeek V4 Flash · 0.0137 to 0.091 · +564% · 2026-10-15
- DeepSeek V4 Pro · 0.2088 to 0.95526 · +358% · 2026-10-15
- GLM 5.3 Flash · 0.036 to 0.15 · +317% · 2026-10-15
- GLM 5.3 · 0.03 to 0.079 · +163% · 2026-10-16
- DeepSeek V4 Flash 0731 · 0.0069 to 0.016 · +132% · 2026-10-15
- Qwen3 VL 8B Instruct · 0.117 to 0.25 · +114% · 2026-10-15
- GPT-5.6 Sol Pro (batch) · 1 to 2 · +100% · 2026-10-15
Provider price spreads
Input price per M tokens, cheapest versus priciest provider:
- GLM 5.3 · 37 providers · Morph 0.04 vs Alibaba 2.8 · spread 6900%
- GLM 5.3 Flash · 30 providers · OpenInference 0.048 vs NextBit 0.25 · spread 421%
- DeepSeek V4.1 Flash · 29 providers · OpenInference 0.0464 vs BaseTen 0.6 · spread 1193%
Benchmark results added to BenchGecko
A selection of the 29 models with results added this week. Some results may come from earlier tests.
- Claude Haiku 5.5 · Artificial Analysis · GDPval 55.9 · Artificial Analysis · Humanity's Last Exam 44.4 · FrontierMath-Tier-4-v2-Private 46.3 · GPQA diamond 86.2
- Gemini 4 Argon · SimpleQA Verified 76 · Artificial Analysis · Quality Index 52.6 · OTIS Mock AIME 2024-2025 100 · Artificial Analysis · CritPt 27.1
- Mistral Large 4 · GPQA diamond 84.1 · Proofbench 10 · Mystery Game Puzzles 8.6 · SimpleQA Verified 20
- DeepSeek V4.1 Flash · FrontierMath-Tiers-1-3-v2-Private 67.4 · ARC-AGI 94.5 · APEX-Agents 39.5 · OTIS Mock AIME 2024-2025 98.3
- Step 5 Preview · Artificial Analysis · Quality Index 43.7 · Artificial Analysis · MMMU Pro 76.4 · Artificial Analysis · Humanity's Last Exam 46.5 · Artificial Analysis · SciCode 58.9
- Nemotron 3.5 Super VL · Artificial Analysis · CritPt 10.6 · Artificial Analysis · Humanity's Last Exam 30.7 · Artificial Analysis · GDPval 33.1 · Artificial Analysis · Long Context Reasoning 69
Mindshare
Current share by model family. Week-over-week change is not available.
- Qwen · 23%
- GLM · 16.5%
- Gemini 4 · 15%
- DeepSeek V4 · 8.3%
- GPT-6 · 8%
- Gemma · 7.1%
- Claude Opus · 5.3%
- Gemini 3 · 5.3%
Company updates
- 2026-10-14 · Labelbox valuation now $1.3B
- 2026-10-14 · Replit headcount now 251
- 2026-10-15 · Micron Technology reported fiscal-year revenue of $133.2B
Indices
Values as of 2026-10-19. 7-day change is not available.
- Gecko Pulse · 37
- Bubble Index · 274
- Mindshare Pulse · 40
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.