AI Agents Weekly: GPT-5.3-Codex-Spark & More — Summary

From Elvis Saravia's AI Newsletter, February 14, 2026

Main Thesis

This issue covers a packed week in AI agents and frontier models, headlined by OpenAI's new agentic coding model, Zhipu AI's open-source powerhouse, and a wave of breakthroughs across autonomous systems, benchmarks, and developer tooling.

Key Stories (Accessible Content)

🔥 GPT-5.3-Codex-Spark (OpenAI)

OpenAI's most capable agentic coding model to date, running 25% faster than its predecessor.
Self-developing: Early versions of GPT-5.3 were used to debug its own training, manage deployment, and interpret evaluation results — making it the first OpenAI model instrumental in its own creation.
Beyond coding: Handles professional knowledge-work outputs including presentations, spreadsheets, and documentation. Wins or ties 70.9% of evaluations on the GDPval knowledge-work benchmark.
Cybersecurity flag: First OpenAI model to hit "high" cybersecurity capability under their Preparedness Framework — meaning it could meaningfully enable real-world cyber harm if misused. OpenAI responded by announcing a $10M API credits program for cyber defense research.

🧠 GLM-5 (Zhipu AI)

A massive 744B-parameter Mixture-of-Experts (MoE) model with 40B active parameters, built specifically for agentic intelligence and multi-step reasoning.
Hardware independence: Trained entirely on Huawei Ascend chips using the MindSpore framework — no US-manufactured semiconductors involved.
Agent Mode: Native autonomous task decomposition, breaking high-level goals into subtasks with minimal human input. Can convert raw prompts into polished .docx, .pdf, and .xlsx documents.
Training scale: Pre-trained on 28.5 trillion tokens (a 23.9% increase over GLM-4.7). Uses a novel RL technique achieving record-low hallucination rates.
Open & affordable: Released under MIT license with open weights. Available on OpenRouter at ~$0.80/M input tokens and $2.56/M output tokens — roughly 6× cheaper than comparable proprietary models.

Other Headlines (Paywalled — Titles Only)

MiniMax M2.5 — New open-source model drop
Recursive Language Models — Replacing context stuffing
OpenAI ships 1M lines with zero manual code
Agentica pushes ARC-AGI-2 with recursive agents
Chrome WebMCP early preview launched
Anthropic raises $30B at a $380B valuation
Excalidraw launches official MCP server
Hive agent framework evolves at runtime
Waymo begins 6th-gen autonomous operations
Gemini 3 Deep Think solves 18 open mathematical problems

Practical Takeaways

Agentic coding is maturing fast — GPT-5.3-Codex-Spark sets a new bar for autonomous software development, including self-referential model improvement.
Open-source is competitive — GLM-5 challenges proprietary frontier models at a fraction of the cost, with full hardware sovereignty.
Cybersecurity risk is real — As models hit "high" capability thresholds, responsible deployment frameworks and defense investment are becoming non-negotiable.
Agent infrastructure is exploding — MCP servers, agentic frameworks, and recursive agent architectures are rapidly becoming standard developer tooling.
Hardware geopolitics matter — GLM-5's Huawei Ascend training stack signals a maturing alternative AI hardware ecosystem outside US supply chains.

Note: No arXiv papers were linked or cited in the accessible portion of this article.

Infographic