This week in AI infra
Every deprecation, API change, pricing shift, and model update that could affect builders shipping on AI APIs -- risk-classified and ready to share.
High-impact changes
24 changes across 5 providers.
Qwen3.6-27B-NVFP4 momentum +1%
nvidia model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
🆕 Custom agents in GitHub Copilot CLI. Define roles, tools, and guardrails in Markdown, then run consistent workflows for security audits,
🆕 Custom agents in GitHub Copilot CLI. Define roles, tools, and guardrails in Markdown, then run consistent workflows for security audits, release notes, incident response, and more. https://t.co/5RDIEzwdHl
Kimi K2.7 Code is the first open-weight model you can select in the GitHub Copilot model picker. What does that mean for you? @burkeholland
Kimi K2.7 Code is the first open-weight model you can select in the GitHub Copilot model picker. What does that mean for you? @burkeholland explains how this low-cost, high-performance model gives you more choice and flexibility in your workflow. ▶️ https://t.co/rxkmT2cABP
🆕 @Kimi_Moonshot's Kimi K2.7 Code is now generally available in GitHub Copilot. This is the first open-weight model offered as a selectable
🆕 @Kimi_Moonshot's Kimi K2.7 Code is now generally available in GitHub Copilot. This is the first open-weight model offered as a selectable option in the Copilot model picker. 🎉 Early testing shows Kimi K2.7 is a lower-cost option with strong performance comparable to highly https://t.co/jHXIwEQU51
AI is shifting from model training to always-on token production, and that shift demands a new business model. NVIDIA is partnering with AI
AI is shifting from model training to always-on token production, and that shift demands a new business model. NVIDIA is partnering with AI clouds to deploy large‑scale, multi‑tenant AI factories through revenue-sharing and credit-support. This opens up compute access to the https://t.co/PSQdTQJTga
OpenAI may cut AI subscription prices as Anthropic rivalry intensifies: Report
OpenAI is reportedly exploring lower subscription and API pricing as rising AI costs and intensifying competition with Anthropic reshape the generative AI market.
Action: Recompute cost guardrails and routing thresholds.
How to Use the DeepSeek API for Free (or Near-Free)
DeepSeek has become one of the most talked-about names in AI for a simple reason: its models deliver strong reasoning and coding at some of the lowest prices in the market. Naturally, the next question every developer asks is whether
Action: Validate API compatibility and update integration tests.
Collin Hogue-Spears Announces Release of From Lab to Life: How AI Works in China
Collin Hogue-Spears announces the August 4, 2026 publication of From Lab to Life: How AI Works in China through Gatekeeper Press.
Action: Benchmark candidate model behavior before adopting in production.
Agents-A1 momentum +30%
InternScience model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
Qwythos-9B-Claude-Mythos-5-1M-GGUF momentum +1%
empero-ai model showing momentum in AI Model.
Action: Run model migration checks for quality, latency, and cost.
GLM-5.2 momentum +16%
zai-org model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
Unlimited-OCR momentum +2%
baidu model showing momentum in AI Model.
Action: Run model migration checks for quality, latency, and cost.
Ornith-1.0-35B-GGUF momentum +2%
deepreinforce-ai model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF momentum +3%
yuxinlu1 model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
DeepSeek-V4-Pro-DSpark momentum +30%
deepseek-ai model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
Ornith-1.0-9B momentum +5%
deepreinforce-ai model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
Ornith-1.0-9B-GGUF momentum +1%
deepreinforce-ai model showing momentum in LLM.
Action: Run model migration checks for quality, latency, and cost.
Evolution of AI: From Turing Test to ChatGPT; How Years of Innovation Created Multi-Trillion-Dollar Economy
The decisive break came in late 2022. As per TECHi's AI timeline, citing Reuters reporting, ChatGPT hit 100 million monthly active users within roughly two months of its November 2022 launch — a milestone that took TikTok nine months and Instagram two and a half years to achieve. The effect on capital markets was immediate: according to the same source, venture capital funding for AI startups surg
Action: Benchmark candidate model behavior before adopting in production.
StarTimes deepens investment in local content, AI innovation with launch of Sura ya Pili
By MKT Correspondent StarTimes Media has reinforced its commitment to Kenya’s creative industry with the launch of Sura The post StarTimes deepens investment in local content, AI innovation with launch of Sura ya Pili appeared first on The Mt Kenya Times .
Action: Benchmark candidate model behavior before adopting in production.
Who owns intelligence?
Artificial intelligence (AI) has largely been discussed through the prism of technological supremacy: which company has the most powerful model, who is winning the benchmark race, and how quickly AI will transform jobs and industries. Yet, a recent exchange between Palantir CEO Alex Karp and Zoho’s Sridhar Vembu points to a more fundamental issue. The real contest may not be over building the best AI models but over owning the economic value they create. Karp’s contention that enterprises must own the “means of production” in the AI era, a view endorsed by Vembu, shifts the debate from technology to economics.
Action: Benchmark candidate model behavior before adopting in production.
IIT-Mandi develops AI framework to speed up molecular analysis, biomedical diagnostics
Researchers at the Indian Institute of Technology (IIT), Mandi, have developed BioFASTNet (Biomedical Fragmented Attention Spectral Transformer Network), an advanced artificial intelligence framework that enables rapid, accurate and automated interpretation of Fourier Transform Infrared (FTIR) spectra. The innovation aims to simplify molecular analysis by eliminating the need for extensive preprocessing and expert-driven interpretation, paving the [...]
Action: Re-run latency/cost checks and adjust timeout budgets.
Tadepalli techie’s AI paper selected for International Conference on Machine Learning in Seoul
This specialised AI model automates the review of lawyer invoices to flag line-item errors and overbilling.
Action: Benchmark candidate model behavior before adopting in production.
HCLTech wins $1.14 billion AI contract from European Fortune Global 50 company
HCLTech has won a strategic contract worth $1.14 billion from a Europe-headquartered Fortune Global 50 company to modernise the client’s digital workplace and enterprise network operations, the Indian IT services company said on Friday. The agreement is one of the largest deals announced by an Indian IT services company this year. The company said the contract will involve establishing an artificial intelligence-driven operating model to transform and manage the customer’s global digital workplace and enterprise networks. HCLTech did not identify the client.
Action: Benchmark candidate model behavior before adopting in production.
Kotak names TCS as top large cap pick but warns Infosys, Wipro face weak Q1 – Prefer 3 midcap IT stocks
Artificial intelligence is beginning to redraw the competitive landscape for India’s IT services industry, but Kotak believes the biggest winners may not be the traditional large-cap players. Instead, the brokerage said select mid-tier companies with stronger execution, healthy deal momentum and differentiated capabilities are better positioned to navigate an environment marked by AI-led pricing pressure, slowing discretionary spending and cautious client budgets.
Action: Recompute cost guardrails and routing thresholds.