Telexed

telexed ~ home★4 and up · hourly · UTC+09LIVE

TELEXED// solo-operator signal radar · Issue 843

AI news through a solo-operator lens — only what changes your day5 of 843

FILTER[All][Agents & tools][Models & API][Generative media][Infra & SaaS][ASO & growth][Indie business][Idea signals][Other][★6+ high-signal]

GeekNews ✕clear filters

Fri, May 291 dispatches

#0843
#0843Models & API GeekNews2 days ago
Anthropic Ships `Claude Opus 4.8`
90radar
Same-price upgrade over 4.7, with better benchmarks and collaboration. Worth re-testing as a coding and agent default before changing spend assumptions.
- Built on Claude Opus 4.7; this is an incremental top-tier model refresh, not a new product line.
- Benchmark performance improved while pricing stays unchanged. Existing API cost models do not need recalculation first.
- Collaboration capability was strengthened, so agent workflows and pair-programming style tasks deserve fresh evals.
Source: news.hada.io/topic?id=29960Read original →
FIG-8431:1
90radar
FIG-8431:1

Tue, May 261 dispatches

#0842
#0842Models & API GeekNews5 days ago
`DeepSeek` cuts `V4-Pro` API pricing by 75% permanently
100radar
V4-ProLLM API model — DeepSeek's flagship commercial model
A permanent cut changes the cost floor for hosted LLM APIs. Use it as leverage for inference-heavy features and vendor price checks.
- The reduction is 75% and permanent, not a promo. V4-Pro can now reset budget assumptions for high-volume inference.
- This directly pressures OpenAI, Anthropic, and Google-style API pricing. Keep abstraction layers ready for model swaps.
- Cheaper calls make batch jobs, agents, and summarization pipelines easier to ship, but quality and latency still need testing.
Source: news.hada.io/topic?id=29882Read original →
FIG-8421:1
100radar
FIG-8421:1

Sat, May 231 dispatches

#0841
#0841Models & API GeekNewslast week
`DeepSeek-V4-Pro` Makes Discounted API Pricing Permanent
100radar
DeepSeek-V4-ProLLM API model — permanent one-quarter pricing
Promo pricing is now the baseline. Token-heavy features get a cheaper model option, so cost tests deserve another pass.
- The 75% discount stays after the promo, keeping DeepSeek-V4-Pro at one-quarter of the prior price.
- Billing is based on 1M-token rates, with input and output usage deducted directly from balance.
- This changes the math for summarization, batch analysis, and agent loops where output volume can dominate cost.
Source: news.hada.io/topic?id=29797Read original →
FIG-8411:1
100radar
FIG-8411:1

Thu, May 211 dispatches

#0840
#0840Models & API GeekNewslast week
`Qwen3.7-Max`: Agent-First Proprietary Model
70radar
Qwen3.7-MaxProprietary LLM — built for long agent runs
A proprietary model is being positioned for coding, office automation, and very long autonomous runs. Strong benchmark numbers make it worth testing for agent workflows, though API cost and access still decide adoption.
- Targets coding, debugging, office automation, and hundreds to thousands of autonomous steps; this is agent runtime territory, not simple chat.
- Scores 69.7 on Terminal Bench 2.0-Terminus and 92.4 on GPQA Diamond; useful signal for coding plus reasoning evals.
- The reported 35-hour autonomous run matters for long workflows, but real value depends on reliability, tool use, and pricing.
Source: news.hada.io/topic?id=29716Read original →
FIG-8401:1
70radar
FIG-8401:1

Wed, May 201 dispatches

#0839
#0839Models & API GeekNews2 weeks ago
`Gemini 3.5 Flash` targets long-running agents and coding
100radar
Google is pushing the fast tier into frontier-agent territory. Recheck Gemini automation stacks where latency matters and quality was the blocker.
- First Gemini 3.5 model combines frontier-level intelligence with execution ability, aimed at long-running agent and coding tasks.
- Keeps the Flash-series speed profile, so it competes for production flows where Pro-class latency was too expensive.
- Scores 76.2% on Terminal-Bench 2.1 and 1656 Elo on GDPval-AA, beating Gemini 3.1 Pro in the cited benchmarks.
Source: news.hada.io/topic?id=29670Read original →
FIG-8391:1
100radar
FIG-8391:1