[KongchangAI]
ModelAxion 1.0-35B / Axion1.0

Axion 1.0

A 35B-parameter fully open-source agent foundation model released by ByteDance. It employs a minimalist two-stage training recipe of SFT plus reinforcement learning, skipping mid-training, DPO, and OPD, and emphasizes verifiable rewards and tool-calling capabilities.

Timeline (last 90 days)

Nov 4

The ByteDance team released Axion 1.0, a fully open 35B-parameter agent foundation model, and simultaneously released the paper, code, datasets and project homepage.

Unverified50%
Nov 4

Axion 1.0's training recipe has only three stages: two lightweight SFT stages and one reinforcement learning stage.

Unverified50%
Nov 4

Axion 1.0's training pipeline does not include dedicated mid-training, direct preference optimization (DPO) or on-policy distillation (OPD).

Unverified50%
Nov 4

The first SFT stage adopts a "model-aware data construction" method: first synthesizing challenging problems, then adversarially strengthening these problems based on the base model's own successful solving strategies.

Unverified50%
Nov 4

The second SFT stage built a scalable data pipeline for constructing diverse and automatically verifiable tasks, based on real data sources and deterministic computation.

Unverified50%
Nov 4

The reinforcement learning stage aims to further strengthen long-horizon reasoning ability as well as the model's ability to interact with external tools and environments.

Unverified50%
Nov 4

Officials say Axion 1.0 achieved strong performance on a broad set of agent benchmarks and is competitive with leading commercial models on several of them.

Unverified50%
Nov 4

The official materials do not provide specific benchmark scores or comparison figures.

Unverified50%
Nov 4

Axion 1.0's openness extends beyond the model weights, with the complete recipe released as well, including data construction, generation, filtering, validation, prompts, and training code.

Unverified50%

All Facts (9)

Unverified

Axion 1.0's openness extends beyond the model weights, with the complete recipe released as well, including data construction, generation, filtering, validation, prompts, and training code.

50%
Unverified

Axion 1.0's training recipe has only three stages: two lightweight SFT stages and one reinforcement learning stage.

50%
Unverified

Axion 1.0's training pipeline does not include dedicated mid-training, direct preference optimization (DPO) or on-policy distillation (OPD).

50%
Unverified

The first SFT stage adopts a "model-aware data construction" method: first synthesizing challenging problems, then adversarially strengthening these problems based on the base model's own successful solving strategies.

50%
Unverified

The second SFT stage built a scalable data pipeline for constructing diverse and automatically verifiable tasks, based on real data sources and deterministic computation.

50%
Unverified

The reinforcement learning stage aims to further strengthen long-horizon reasoning ability as well as the model's ability to interact with external tools and environments.

50%
Unverified

Officials say Axion 1.0 achieved strong performance on a broad set of agent benchmarks and is competitive with leading commercial models on several of them.

50%
Unverified

The official materials do not provide specific benchmark scores or comparison figures.

50%
Unverified

The ByteDance team released Axion 1.0, a fully open 35B-parameter agent foundation model, and simultaneously released the paper, code, datasets and project homepage.

50%

Source Articles