# AI Coding: Half-Price Terra Enters; Opus 5.5 Holds Lead

> Claude Opus 5.5 keeps the lead across intelligence and coding agents on Nov. 1. OpenAI added GPT-6.1 Terra, a half-price model built for computer use.

- URL: https://roobrief.com/en/posts/ai-weekly-coding-models-nov-01
- Language: en
- Published: 2026-11-01T14:20:00.000Z
- Tags: ai
- Source: https://artificialanalysis.ai/leaderboards/models

**Claude Opus 5.5 retained top honors across overall intelligence and coding agents, while OpenAI expanded its value lineup with GPT-6.1 Terra, a half-price model tailored for computer use.**

This is a weekly brief for choosing models across coding and agentic workflows. Rankings are based on Artificial Analysis (AA) leaderboards, with figures taken from a snapshot around Nov. 1, 2026.

The Intelligence Index evaluates overall capability across math, science, coding, and reasoning. The Coding Agent Index measures performance on realistic open-source repository tasks, interactive terminal operations, and technical software Q&A. "Hot models" is an editorial roundup of new launches and default changes over the past week rather than an official rank.

## Intelligence Index top 3 and steady frontier

AA Intelligence Index v4.3.2 evaluates models per configuration setting. The table below lists each model's best setting; second place remains tied at 56:

| Rank | Model (setting) | Score | Cost per task |
|---|---|---|---|
| 1 | Claude Opus 5.5 (max, fallback) | 58 | $5.98 |
| 2 | Claude Sonnet 5.5 (max, fallback) | 56 | $5.46 |
| 2 | Muse Wonder Full (max) | 56 | not listed |

The top three standings stayed level with last week. Moonshot AI's Kimi K3.1 Pro—shipped on Oct. 23—held fourth place with an Intelligence Index score of 55 and an estimated cost of $2.11 per task, remaining the highest-scoring model among open-weight(releasing model weights so others can download, run, or adapt them) contenders. Opus 5.5 launched Sept. 22 and Sonnet 5.5 on Sept. 28; both feature a 1-million-token context window. A token(a small piece of text, roughly part of a word, that AI models read and write) costs $4 input and $20 output per million tokens on Opus, compared to $2 input and $10 output on Sonnet. Kimi K3.1 Pro matches Sonnet on headline token rates ($2.00 input, $10.00 output), but higher average output generation pushes its benchmark task cost to $2.11.

## Coding Agent Index v2.0 settles with Kimi listed

The Coding Agent Index v2.0 overhaul launched on Oct. 23 is now in its second week, and the frontier hierarchy has stabilized. Evaluated across 290 total workloads—comprising Cognition's FrontierCode v1.1 (100 repository tasks), Terminal-Bench 4.0 (66 tasks), and SWE-Atlas-QnA-AA (124 tasks)—Claude Opus 5.5 on Claude Code held firmly onto first place at 61.7 points.

| Rank | Model and harness | v2.0 score | Notes |
|---|---|---|---|
| 1 | Claude Code + Claude Opus 5.5 (max) | 61.7 | ~$11.42 per task |
| 2 | Claude Code + Claude Sonnet 5.5 (max) | 59.3 | Consistent upper tier |
| 3 | Muse Code + Muse Wonder Full (max) | 58.9 | Subscription only |
| 4 | Claude Fable 5.1 (max, fallback) | 57.6 | Upper tier baseline |
| 5 | Codex + GPT-6.1 Sol (max) | 54.1 | ~$1.50 per task |
| 6 | Kimi Code + Kimi K3.1 Pro (max) | 52.4 | ~$2.80 per task, official entry |

The standout addition this week is Moonshot AI's official harness pairing, Kimi Code + Kimi K3.1 Pro, which entered the board at 52.4 points in sixth place. This marks the strongest performance on the v2.0 agent board among models with upcoming open weights. While Opus's deep inference(using a finished model to produce answers) capacity continues to anchor its lead in complex repository navigation and refactoring, Codex with GPT-6.1 Sol held fifth place at 54.1 points (~$1.50 per task). Lower down the rankings, DeepSeek V4.1 Flash reached 47.3 points (~$0.39 per task), and Gemini 4 Argon on Antigravity CLI remained in the low-to-mid 40s.

## Value top 3 and Terra entry

AA does not maintain a dedicated "value top 3." The table below reflects an editorial comparison of Intelligence Index performance against per-task expenditure. Because Sol's scores vary by reasoning effort, max represents the upper tier and medium the lower-cost counterpart:

| Rank | Model (setting) | Intelligence Index | Cost per task |
|---|---|---|---|
| 1 | GPT-6.1 Sol (medium) | 48 | $0.21 |
| 2 | GPT-6.1 Terra (max) | 47 | $0.35 |
| 3 | GPT-6.1 Sol (max) | 52 | $0.72 |

The headline movement on the value board is the debut of OpenAI's GPT-6.1 Terra, released on Oct. 26. Terra scored 47 on the Intelligence Index at $0.35 per task, securing the second spot. Priced at $1 per million input tokens and $5 per million output tokens—half the rate of Sol—Terra significantly lowers operating costs for agent pipelines heavy on visual analysis and screen interaction.

In the higher-intelligence tier, Kimi K3.1 Pro (intelligence 55 at $2.11) continues to offer a viable alternative at less than half the per-task expenditure of Opus ($5.98) or Sonnet ($5.46). For minimal-cost operations, MiMo-V2.6-Pro (intelligence 46 at $0.13) and Claude Haiku 5.5 (max, intelligence 43 at $0.21) maintain their strong baselines.

## Hot models last week, and what shipped new

Four notable developments unfolded across coding and agent infrastructure between Oct. 26 and Nov. 1:

1. **OpenAI GPT-6.1 Terra launch** — released Oct. 26. A mid-tier model tuned for computer use (the capability to perceive UI screens and control them with a mouse and keyboard) and visual reasoning. It cuts API pricing in half compared to Sol ($1 input, $5 output per million tokens) with a 1.1-million-token context window, shipping simultaneously on the OpenAI API, ChatGPT paid tiers, GitHub Copilot, and Amazon Bedrock.
2. **Kimi K3.1 Pro Coding Agent v2.0 listing** — posted Oct. 28. Following its 55 intelligence score, Kimi K3.1 Pro logged 52.4 points on the Coding Agent Index v2.0, establishing the top mark for open-weight contenders.
3. **NVIDIA Nemotron 3.5 Bolt E4B on-device optimization** — released Oct. 29. Tuned for 4GB VRAM(memory dedicated to a graphics card) targets, updated local runtime optimizations and Jetson device deployment guides became available for running lightweight agent tools locally.
4. **DeepSeek-V4.1-Flash-1017 free server tool promotion ended** — Nov. 1. The two-week free tier for Search & Fetch server tools that accompanied the Oct. 17 release expired at 16:00 UTC on Oct. 31 (01:00 KST Nov. 1), moving to standard metered billing.

New open and local weights released across the week:

| Date | Model / Product | Developer access path |
|---|---|---|
| 10/26 | GPT-6.1 Terra | OpenAI API, ChatGPT, GitHub Copilot, Bedrock |
| 10/28 | Kimi K3.1 Pro (Agent v2.0 listed) | Kimi Code, Moonshot API |
| 10/29 | Nemotron 3.5 Bolt E4B optimization | Local runtimes, Jetson deployment recipes |
| 11/01 | DeepSeek V4.1 Flash server tools paid | DeepSeek platform API |

## Common misconceptions

While GPT-6.1 Terra operates 3x faster on screen manipulation and costs half as much as Sol, pure algorithm design and deep code refactoring remain superior on Sol and Astra. Developers should not treat Terra as an across-the-board replacement for all coding tasks. The end of DeepSeek Flash-1017's free promotion applies only to auxiliary server tools like web search, not the model's underlying token pricing. Lastly, the 61.7 score on the Coding Agent Index reflects the full Claude Code harness running with max effort, not raw Opus 5.5 model weights in isolation.

## What to watch next

- Agent workflow adoption shifts as Gemini 4 Argon's introductory API pricing ($2/$10) enters its final month before expiring Nov. 30
- Formal open-weights release dates for Kimi K3.1 Pro and local inference framework optimizations
- Enterprise adoption pace of GPT-6.1 Terra for browser automation and screen-level computer use pipelines
- Ecosystem growth of local and on-device agent tools using lightweight models like Nemotron Bolt E4B

## Sources

- [Artificial Analysis Intelligence Index](https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index) (Artificial Analysis)
- [Coding Agent Benchmarks](https://artificialanalysis.ai/agents/coding-agents) (Artificial Analysis)
- [LLM Leaderboard](https://artificialanalysis.ai/leaderboards/models) (Artificial Analysis)
- [Introducing GPT-6.1 Terra](https://openai.com/index/introducing-gpt-6-1-terra/) (OpenAI)
- [Moonshot AI Kimi K3.1 Pro Announcement](https://kimi.ai/blog/k3-1-pro) (Moonshot AI)
- [DeepSeek-V4.1-Flash-1017](https://api-docs.deepseek.com/news/news261017/) (DeepSeek)
- [Gemini 4 Argon is now available for everyone](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/) (Google)
- [Introducing Claude Opus 5.5](https://www.anthropic.com/claude-opus-5-5) (Anthropic)
- [Introducing GPT-6.1 Sol](https://openai.com/index/introducing-gpt-6-1-sol/) (OpenAI)