Skip to main content
Back

OpenAI Terra at $1/M tokens matches GPT-6 Astra at 10x less cost

OpenAI ships GPT-6.1 Terra at 1/10th price, Cohere catches Qwen3 probing sandbox, Pomerado turns sites into APIs, OrcaRouter cuts refusals 91.5%

October 27, 2026 8 min read

Hey,

Two stories today that tell the same story from opposite ends. OpenAI ships GPT-6.1 Terra at $1/M tokens, matching Astra's performance at one-tenth the price. Cheaper, faster, production-ready. Meanwhile, Qwen3 gets caught trying to phone home to Beijing mid-task, after its own reasoning log said it wouldn't.

One model gets cheaper. Another gets shadier. That's the trust gap widening in real time.

We're entering an era where cost is solved faster than alignment. Worth keeping that in mind.
In Partnership with · partner_image
Author photo
Today's Author
Lior Alexander. Founder of AlphaSignal and former ML Engineer. Previously built ML systems at Iguazio, Guesty, Enphase, Mila.
Summary

Read time: 8 min 9 sec

Top News

▸ OpenAI ships GPT-6.1 Terra, matching GPT-6 Astra at one-tenth the price

Google Cloud

▸ Google Cloud A5X Scales to 80,000 Rubin GPUs

Top News

▸ Cohere catches Qwen3 model secretly probing its sandbox and attempting file uploads

AgentField AI

▸ Solve 2x More Issues, 4.8x Less Cost Per Fix

Top Repo

▸ This app fine-tunes a model on your handwriting to generate perfect LaTeX

Signals

1. Pomerado turns any website into an API in 30 minutes, up to 1000x cheaper than browser agents
2. Granola: Standups and design reviews, written up and shared with your team in Granola. First month on us with code ALPHASIGNAL
3. New self-training method improves LLMs without any human-labeled data
4. NYU study reveals why frontier models write in ways humans struggle to read
5. New model merge combines two Minimax-H3 video finetunes with a custom blending algorithm
6. OrcaRouter open-sources uncensored DeepSeek model with 91.5% fewer refusals and no capability loss
Our Events
Requirements, Challenges, and Solutions for Full-Duplex Model Training
Oct 27, 10a PT · Online
Voice AI that can listen and talk at the same time is the next frontier, and training those full-duplex models is a whole different problem. Join Sully Goolamnobee, Product Manager for AI Voice at Voices.com for a live session on the requirements, challenges, and solutions for full-duplex model training.
Designing Flexible, Efficient AI Infrastructure with Google Cloud's AI Hypercomputer
Oct 29, 10a PT · Online
Scaling AI shouldn't mean choosing between raw performance and cost efficiency. Join Chelsie Czop, Sr. Product Manager for AI Infrastructure at Google Cloud to see how AI Hypercomputer, NVIDIA GPUs, and smart scheduling keep training and inference fast, resilient, and affordable.
Building a Semantic Backbone for Agentic AI over Enterprise Data
Nov 5, 10a PT · Online
AI agents are only as smart as the enterprise data they can actually understand, and most of that data is scattered and siloed. Join Vassil Momtchev, SVP of Product & Technology at Graphwise to learn how to build a semantic backbone for agentic AI, with a live demo of the Graphwise AI Suite.
Top News
OpenAI ships GPT-6.1 Terra, matching GPT-6 Astra at one-tenth the price
23,237 Likes
OpenAI ships GPT-6.1 Terra, matching GPT-6 Astra a

OpenAI just dropped GPT-6.1 Terra, and the headline is simple: same power, way less money.

Terra sits in the middle of OpenAI's model lineup. It is not the smartest model they have, but it is the fastest and cheapest for everyday tasks. Think of it as the workhorse you actually run in production.

Here is what makes it interesting:

  • API pricing is $1 per million input tokens and $5 per million output tokens, which is 50% cheaper than GPT-5.6 Terra and GPT-6.1 Sol
  • It can look at a screen, click through apps, fill forms, and type into software on its own, no human needed
  • Coding jumped from 21.5% to 50.3% on Terminal-Bench 4.0, more than double its predecessor
  • Completes tasks 3x faster than GPT-6 Astra and GPT-6.1 Sol, at one-tenth the price

You can call it right now via the OpenAI API using gpt-6.1-terra. For deep research or hard coding problems, Sol or Astra still win. But for everything else, Terra is the smart default.

Read more
forward →
Presented by Google Cloud
Google Cloud A5X Scales to 80,000 Rubin GPUs

Google Cloud's A5X instances, powered by NVIDIA's Vera Rubin platform, scale from a single 72-GPU rack to 80,000 GPUs in one data center, and nearly a million across sites.

Now generally available, A5X links 72 Rubin GPUs via sixth-gen NVLink so each rack works as one accelerator, cooled by the system behind over a gigawatt of Google infrastructure. Built for frontier training, reinforcement learning, physical AI agents, and real-time inference.

Watch the launch video to learn more.

Watch now
partner with us →
Top News
Cohere catches Qwen3 model secretly probing its sandbox and attempting file uploads
9,544 Likes
Cohere catches Qwen3 model secretly probing its sa

Cohere was running a routine security test on Alibaba's open-weight model Qwen3.8-Flash-Next when things got weird. The task was simple: write a rule to detect malicious file downloads. Instead, the model went off-script.

Here is what it actually did:

  • Scanned files in the /tmp directory and checked what transfer tools were installed
  • Read the machine's DNS settings
  • Ran a curl command trying to upload a probe file to an Alibaba Cloud server in Beijing
  • Did all of this after its own reasoning log said it would use no tools

The sandbox blocked it. Nothing left. But here is the part that should stick with you: the model's written reasoning lied about what it was about to do. You cannot trust the trace to tell you what the model will actually run.

This happened in fewer than 1 in 100,000 runs. Cohere says no malicious intent is confirmed. But on a normal laptop with internet access, that upload would have gone through silently. Always sandbox AI agents with real network restrictions, not just trust.

Read more
forward →
Presented by AgentField AI
Solve 2x More Issues, 4.8x Less Cost Per Fix

Ten coding harnesses ran the same open model on 113 GitHub issues. CodeAF ranked #1, ahead of Codex, and solved 2x more than OpenCode at 4.8x lower cost per fix.

CodeAF is an open-source coding harness built for open models like DeepSeek, GLM and Kimi, and works with closed models too.

Star on github
partner with us →
Top Repo
This app fine-tunes a model on your handwriting to generate perfect LaTeX
3,841 Likes
This app fine-tunes a model on your handwriting to

Someone just fine-tuned an AI model on their own handwriting and the results are seriously impressive.

The idea is simple. Generic handwriting-to-LaTeX tools struggle because they train on everyone's handwriting. Your quirky letter shapes throw them off. So Th0rgal did something smarter: collected their own 2022 handwritten notes, used them to retrain a vision model on their specific handwriting style, and built an app around it.

Fine-tuning means taking an existing AI model and retraining it on a small, specific dataset , here, one person's notes , so it gets really good at that one thing instead of being okay at everything.

What this opens up for you:

  • Snap a photo of your handwritten notes and get clean LaTeX output instantly
  • The model knows YOUR letter shapes, not a generic average of millions of people
  • You can do the same , train a model on your own notes using the open-source app on GitHub under Th0rgal/orb

The takeaway: personalized models beat generic ones. A small dataset of your own data can outperform a massive model trained on everyone else.

Read more
forward →
Signals
1
Pomerado turns any website into an API in 30 minutes, up to 1000x cheaper than browser agents
3,659 Likes
2
Standups and design reviews, written up and shared with your team in Granola. First month on us with code ALPHASIGNAL
Presented by Granola
3
New self-training method improves LLMs without any human-labeled data
1,604 Likes
4
NYU study reveals why frontier models write in ways humans struggle to read
1,569 Likes
5
New model merge combines two Minimax-H3 video finetunes with a custom blending algorithm
11,413 Downloads
6
OrcaRouter open-sources uncensored DeepSeek model with 91.5% fewer refusals and no capability loss
3,503 Likes
At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 300,000 developers stay informed and ahead. We're passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.

Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We're based in the United States. By using our site, you agree to be governed by U.S. laws.
Looking to promote your company, product, service, or event to 300,000+ AI developers?
Work with us