RSS.bz

AI Daily · Oct 12, 2026: OpenAI raises GPT-6.1 Sol default context to 500k

OpenAI raised GPT-6.1 Sol's default context in Codex and Work from 272k to 500k with no long-context fees on subscription usage, while OpenCode offers $80 of DeepSeek V4.1 Flash usage credit.

Showing the latest briefing (Oct 12, 2026). Every day is generated the next morning.

Pick a date 3/2080: 0 briefings
3/2080
SunMonTueWedThuFriSat
12
3456789
10111213141516
17181920212223
24252627282930
31

Dotted dates have a briefing; click one to open it.

  • OpenAI raises GPT-6.1 Sol default context to 500k in Codex and Work

    OpenAI increased the default context length for GPT-6.1 Sol to 500k tokens when used through Codex and Work, up from 272k (about 80% more), and says it will not charge long-context fees for this model when usage is drawn from a subscription.

    @thsottiaux 1 duplicate reports merged

  • OpenCode gives $80 usage credit for DeepSeek V4.1 Flash

    OpenCode's Operation Cheepseek Phase 3 offers $80 of usage for DeepSeek V4.1 Flash.

    @opencode 1 duplicate reports merged

  • Requesty measures Claude Haiku 5.5 speed, cost and adoption gains

    On the Requesty gateway, Haiku 5.5 vs Haiku 4.5 for the same customers and window: 2.9x faster output, 3.8x longer answers, 3.1x more context per request, 44% of calls ending in a tool call (14% on 4.5), 21% of output as thinking (3% on 4.5) and 48% cheaper per request; p50 streaming output speed was 3.3x Haiku 4.5, and 5 days after launch Haiku 5.5 carried 94% of Haiku tokens.

    @RequestyAI 2 duplicate reports merged

  • QwenGyre brings elastic RL to extreme-long-horizon agent training

    Qwen introduces QwenGyre for extreme-long-horizon agent RL, with an Elastic Scheduler that reallocates GPUs between rollout and training without interrupting live executions, reporting up to 1.85x speedup over Colocate and 1.78x over Async, plus a Trajectory Processor for branching execution.

    @Alibaba_Qwen 1 duplicate reports merged

  • Qoder launches Fast mode with sub-third first-response latency

    Qoder released a new fastest mode called Fast; time to first response is under 1/3 of Auto mode, under 1/2 of Performance mode, and about 1/4 of Ultimate mode, with completion time on 10 everyday dev tasks about half that of comparable market fast modes at the same output quality.

    @qoder_ai_ide 1 duplicate reports merged

  • Command Code ships open decision model Agr, cloud sandboxes, and new model integrations

    Command Code's weekly ship list includes Agr, described as its first open decision model, plus cloud sandboxes and ACP support. It also adds Claude Haiku 5.5 (2x in GOAT), Mistral Large 4 (4x in GOAT), a free model called Glyph Cluster, and a Hooks feature demo.

    @MrAhmadAwais

  • Command Code lists GPT 6 Luna and GPT 5.6 Sol/Luna in its lineup

    Command Code, a coding agent for open models, said its GOAT lineup includes three GPT models: GPT 6 Luna, GPT 5.6 Sol and GPT 5.6 Luna.

    @CommandCodeAI