ring-2.6-1t

Ring-2.6-1t belongs to Ant Group's Ring family of open-source thinking models and shares the same Mixture of Experts architecture as Ling, activating roughly 63 billion parameters per inference out of a total around one trillion. This sparse activation pattern is designed to keep per-query compute manageable while preserving the reasoning capacity expected of a frontier-scale model. The architecture aligns with Ant Group's broader push into trillion-parameter systems tuned for agent-style workflows rather than purely open-ended chat.

According to Ant Group's developer documentation, Ring is explicitly optimized for coding agents, multi-tool collaboration, and long-horizon task execution, and the team reports leading results on agent-focused benchmarks including PinchBench, ClawEval, TAU2-Bench, and GAIA2-search. The 2.6-1T release extends this focus with an adjustable reasoning effort mechanism offering high and xhigh intensity levels, allowing adaptive budget allocation based on task complexity. Combined with a 262K token context window, the model fits well for production agent pipelines, research analysis, and multi-step engineering tasks where consistent task completion matters as much as raw quality.

Requestyring-2.6-1tring

Open provider page Provider docs

Quick Info

Provider: Requesty
Model key: ring-2.6-1t
Release date: May 8, 2026
Last updated: May 8, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.30
Output token cost: $2.50

Limits

Output tokens: 262,144 tokens
Context window: 262,144 tokens

Latest news about ring-2.6-1t

Requesty

Official sourceBenchmark

Requesty's router exposes Inclusion AI's ring-2.6-1t via a Novita AI deployment catalogued at https://www.requesty.ai/models/novita/inclusionai-ring-2.6-1t, with the exact routed model id `novita/inclusionai/ring-2.6-1t`. The page documents that this endpoint was added in May 2026, serves chat completions from US infra Pricing on the Requesty page is set at $0.30 per 1M input tokens and $2.50 per 1M output tokens (8.3x input mark-up), with no prompt caching pricing listed and no per-request markup beyond Requesty's standard 5% pay-as-you-go or 0% on customer keys. The context window is 262K tokens with no max output cap specified, an

Requesty

Coverage

InclusionAI (Ant Group's Ling/Bailing line) released Ring-2.6-1T as a trillion-scale flagship reasoning model, reported on May 9, 2026. The model targets complex production environments such as Agent workflows, engineering development, and scientific analysis, and centers on a configurable Reasoning Effort mechanism in In real-task evaluations cited in the article, the high mode achieved 87.60 on PinchBench, reportedly surpassing GPT-5.4xHigh and Claude-Opus-4.7xhigh, and the xhigh mode scored 95.83 on AIME26 and 88.27 on GPQA Diamond. The piece frames Ring-2.6-1T as shifting frontier competition from parameter scale toward reasoning

Videos about ring-2.6-1t

More models around ring-2.6-1t

Browse family

1 provider

Ring-1T

Currently listed through:

Bailing