The week in AI models.
Sep 28 – Oct 4, 2026. 4 new releases, 9 new leaders on tests, 2 listed price changes and 2 models joining the capability index.
Highlights
Counted from this week’s releases, new leaders and price changes.
Top new releaseGPT-6.1 SolOpenAI, 166.1 on the capability index- Most releasesOpenAI2 new models
Most new top spotsGemini 4 Argon HighNew leader on 6 tests
Biggest price changeDeepSeek V4 Pro+82%, now $0.66 in and $1.98 out
New releases
Models from known makers that the maker sells, several hosts list or a test covers, highest on the capability index first, then the most widely offered.
| Model | Released | Capability index | Price | Context |
|---|---|---|---|---|
| GPT-6.1 SolOpenAI | Sep 29, 2026 | 166.1 | $4.00 | 1.1M |
 Claude Sonnet 5.5Anthropic | Sep 28, 2026 | 165.0 | $4.00 | 1M |
| GPT-6.1 Sol ProOpenAI | Sep 29, 2026 | Not on the index yet | $4.00 | 1.1M |
 Grok Imagine Video 1.5 LitexAI | Oct 1, 2026 | Not on the index yet | No paid listing | 1K |
New leaders
- New leader
Gemini 4 Argon High took the top spot on Arena text at 1525, from Claude Opus 5.5.
- New leaderWithin margin
Gemini 4 Argon High took the top spot on Arena coding at 1559, from Claude Opus 4.6.
- New leaderWithin margin
Gemini 4 Argon High took the top spot on Arena math at 1528, from Claude Fable 5.
- New leader
Gemini 4 Argon High took the top spot on Arena hard prompts at 1550, from Claude Opus 5.5.
- New leaderWithin margin
Gemini 4 Argon High took the top spot on Arena creative writing at 1522, from Claude Opus 5.5.
- New leader
Gemini 4 Argon High took the top spot on Arena instruction following at 1529, from Claude Opus 5.5.
- New leader
GPT-6.1 Sol took the top spot on FrontierMath Tier 4 at 100.0%, from GPT-6 Astra.
- New leader
Claude Opus 5.5 took the top spot on CursorBench at 57.8%, from Claude Fable 5.1.
- New leader
Claude Sonnet 5.5 took the top spot on LiveBench coding at 91.4%, from Claude Opus 5.5.
Price changes
- Price change+82%
DeepSeek’s listed price for DeepSeek V4 Pro went from $0.435 in and $0.87 out to $0.66 in and $1.98 out per 1M tokens.
Price change−38%Alibaba’s listed price for Qwen3.7 Plus went from $0.50 in and $3.00 out to $0.40 in and $1.60 out per 1M tokens.
Joined the capability index
- Joined the index
Claude Opus 5.5 joined the capability index at 167.3.
- Joined the index
Claude Sonnet 5.5 joined the capability index at 165.2.
Releases, prices and context windows from models.dev (MIT); LiveBench (Apache-2.0); Epoch AI (CC BY 4.0); Cursor CursorBench, compiled by Epoch AI (CC BY 4.0); Arena (CC BY 4.0). A week runs Monday to Sunday, UTC. Changes count from when the site began keeping this record, as on What changed. RSS feed