2026-09-26 to 2026-10-02 · Published 2026-10-03
This week in AI · 2026-09-26 to 2026-10-02
Six new models were listed, from Anthropic and OpenAI. BenchGecko recorded 26 price changes, 12 of them drops, and added benchmark results for 198 models.
New models
Six models were listed this week.
- Claude Sonnet 5.5 · Anthropic · 2026-09-29 · $2 input / $10 output per M tokens · 1,000,000 context
- Claude Sonnet 5.5 (batch) · Anthropic · 2026-09-29 · $1 input / $5 output per M tokens · 1,000,000 context
- GPT-6.1 Sol Pro · OpenAI · 2026-09-30 · $2 input / $10 output per M tokens · 1,050,000 context
- GPT-6.1 Sol Pro (batch) · OpenAI · 2026-09-30 · $1 input / $5 output per M tokens · 1,050,000 context
- GPT-6.1 Sol · OpenAI · 2026-09-30 · $2 input / $10 output per M tokens · 1,050,000 context
- GPT-6.1 Sol (batch) · OpenAI · 2026-09-30 · $1 input / $5 output per M tokens · 1,050,000 context
Price changes
26 price changes were recorded, including 12 price drops. The largest input price increases:
- Qwen3.8 27B · $0.0428 to $0.42 per M input tokens · +881% · 2026-09-30
- Kimi K3 · $0.8845 to $3 · +239% · 2026-09-26
- GLM 5.3 Flash · $0.045 to $0.15 · +233% · 2026-09-28
- DeepSeek V4 Pro 0813 · $0.4 to $1.32 · +230% · 2026-09-30
- DeepSeek V4 Flash · $0.04704 to $0.14 · +198% · 2026-09-28
- DeepSeek V4 Pro · $0.348 to $0.95526 · +175% · 2026-09-28
- DeepSeek V3.2 Exp · $0.1344 to $0.27 · +101% · 2026-09-30
- GPT-5.6 Sol Pro · $2 to $4 · +100% · 2026-09-29
Benchmark results added to BenchGecko
Results were added to BenchGecko for 198 models. A sample:
- GPT-5.6 Terra · Dtbench 88.9 · ARC-AGI 96.5 · Balrog 53.2 · Artificial Analysis · Long Context Reasoning 83
- GPT-6 Astra · Frontiercode 53.3 · Mirrorcode 46.7 · WeirdML 93.6 · Furniture Assembly 71.4
- Kimi K3 · Artificial Analysis · MMMU Pro 80.5 · ARC-AGI 94.5 · Deepswe 68.5 · Artificial Analysis · Quality Index 43.6
- Claude Fable 5.1 · WeirdML 92.9 · ARC-AGI 97.5 · Artificial Analysis · Long Context Reasoning 85.3 · Artificial Analysis · CritPt 29.7
- Gemini 3.8 Flash · Dtbench 92.9 · FrontierMath-Tiers-1-3-v2-Private 68.4 · SimpleQA Verified 69.7 · Chess Puzzles 59
- GPT-5.6 Sol · OTIS Mock AIME 2024-2025 100 · PostTrainBench 36.2 · Chess Puzzles 62.1 · Proofbench 83
Mindshare
Share by model family, with the weekly change where available.
Rises:
- GPT-6 · 25.3% · up 553 bps
- GLM · 14.1% · up 523 bps
- Gemini 3 · 4.9% · up 110 bps
- DeepSeek V4 · 6% · up 77 bps
- Gemma · 5% · up 38 bps
Falls:
- Qwen · 18.6% · down
- GPT-5 · 8.6% · down
- Claude Opus · 5.1% · down
Provider price spreads
Input price spread between the cheapest and priciest provider:
- GLM 5.3 · 32 providers · Relace $0.12 to Alibaba $2.8 per M input tokens · spread 2233%
- GLM 5.3 Flash · 32 providers · Relace $0.035 to Cloudflare $0.3 · spread 757%
- DeepSeek V4.1 Flash · 29 providers · Relace $0.02 to Fireworks $0.45 · spread 2150%
Indices
As of 2026-10-03:
- Gecko Pulse · 35
- Bubble Index · 278
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.