Local image understanding, visual question answering, captioning, document and chart analysis, counting, pointing, and multimodal research
Type
Multimodal
Context
4K
Reasoning
6/10
Speed
6/10
Multimodal
Image input
Fine-tuning
Status
Available as an open-weight downloadable checkpoint
Input
No official hosted API pricing; self-hosted weights
Output
No official hosted API pricing; self-hosted weights
View model
→
Local image-and-text understanding, visual question answering, document and chart analysis, image captioning, and multimodal research
Type
Multimodal
Context
4K
Reasoning
6/10
Speed
6/10
Multimodal
Image input
Fine-tuning
Status
Available open-weight checkpoint; preview release
Input
No official hosted API pricing; downloadable weights for self-hosted deployment
Output
No official hosted API pricing; deployment cost depends on self-hosted or third-party infrastructure
View model
→
Self-hosted image understanding, visual question answering, captioning, image grounding, pointing, counting, and multimodal research.
Type
Multimodal
Context
4K
Reasoning
7/10
Speed
2/10
Multimodal
Image input
Fine-tuning
Status
Available open-weight legacy research model; newer Molmo 2 models are Ai2's current successor family.
View model
→
Local image understanding, visual question answering, image captioning, document and chart analysis, counting, and experimentation with open multimodal models
Type
Multimodal
Context
4K
Reasoning
5/10
Speed
8/10
Multimodal
Image input
Status
Available open-weight preview checkpoint
View model
→
Efficient local image and video understanding, visual grounding, pointing, captioning, counting, tracking, and multimodal research
Type
Multimodal
Reasoning
6/10
Speed
8/10
Multimodal
Image input
Video input
Status
Current open-weight model
Input
No official hosted API pricing; self-hosted model weights
Output
No official hosted API pricing; self-hosted model weights
View model
→
Open multimodal research, video question answering, visual grounding, pointing, counting, captioning, object tracking, document understanding, and robotics perception
Type
Multimodal
Context
37K
Reasoning
7/10
Speed
6/10
Multimodal
Image input
Video input
Status
Current open-weight model; downloadable checkpoint
View model
→
Open multimodal research, local image and video understanding, visual grounding, pointing, counting, captioning, and tracking
Type
Multimodal
Context
66K
Reasoning
7/10
Speed
6/10
Multimodal
Image input
Video input
Status
Current open-weight model
Input
Not applicable; open-weight model with no official hosted API token pricing found
Output
Not applicable; open-weight model with no official hosted API token pricing found
View model
→
Open research and downstream fine-tuning for vision-guided robotic manipulation, spatial reasoning, trajectory planning, and robot action prediction
Type
Multimodal
Context
4K
Reasoning
7/10
Speed
5/10
Multimodal
Image input
Media output
Status
Available open-weight research model; superseded by MolmoAct2 as Ai2's newer MolmoAct generation
View model
→
Robotic manipulation research, action reasoning, downstream mid-training, and reproducing zero-shot SimplerEnv experiments
Type
Multimodal
Reasoning
7/10
Speed
4/10
Multimodal
Image input
Fine-tuning
Status
Available open-weight preview checkpoint
Input
No official hosted API pricing; self-hosted checkpoint
Output
No official hosted API pricing; self-hosted checkpoint
View model
→
Open robotics research, visual action reasoning, robot-manipulation experiments, and fine-tuning on custom robot datasets
Type
Multimodal
Reasoning
7/10
Speed
4/10
Multimodal
Image input
Media output
Status
Available open-weight preview checkpoint; intended for fine-tuning and downstream post-training
View model
→
Open robotics research, embodied visual reasoning, robot-policy fine-tuning, and manipulation tasks on supported or closely related hardware
Type
Multimodal
Context
16K
Reasoning
8/10
Speed
8/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight foundation checkpoint
View model
→
Depth-aware robot manipulation research, embodied reasoning, and fine-tuning vision-language-action policies for target robot embodiments.
Type
Robotics
Reasoning
7/10
Speed
5/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight foundation checkpoint
View model
→
Language-guided 3D trajectory forecasting, robotics planning research, and motion-conditioned video generation
Type
Multimodal
Reasoning
2/10
Speed
4/10
Multimodal
Image input
Video input
Status
Documented research variant; no separately released public checkpoint identified in the current first-party model collection
View model
→
Open research and applications requiring image or video grounding, visual pointing, object localization, counting, tracking, spatial reasoning, and multimodal analysis.
Type
Multimodal
Reasoning
7/10
Speed
6/10
Multimodal
Image input
Video input
Status
Current open-weight model
Input
No official hosted API pricing; downloadable checkpoint
Output
No official hosted API pricing; downloadable checkpoint
View model
→
GUI screenshot grounding, screen-element localization, computer-use perception, and research on pointing models
Type
Multimodal
Context
37K
Reasoning
4/10
Speed
5/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight research model
View model
→
Video object pointing, temporal grounding, object counting, tracking, and research on language-guided video understanding
Type
Multimodal
Context
35K
Reasoning
4/10
Speed
5/10
Multimodal
Video input
Status
Current open-weight research model
View model
→
Open research, reproducible language-model experiments, local deployment, English text generation, and fine-tuning
Type
General Purpose
Context
4K
Reasoning
6/10
Speed
4/10
Fine-tuning
Streaming
Status
Available open-weight model; not formally deprecated or retired
View model
→
Local inference, open-model research, benchmarking, continued pretraining, and task-specific fine-tuning.
Type
General Purpose
Context
4K
Reasoning
5/10
Speed
7/10
Fine-tuning
Status
Available as a downloadable open-weight model; no official first-party hosted API identified.
Input
No official Ai2 hosted API price; self-hosted model weights are downloadable.
Output
No official Ai2 hosted API price; self-hosted model weights are downloadable.
View model
→
Fully open language-model research, local inference, reproducible training experiments, evaluation, and downstream fine-tuning
Type
General Purpose
Context
4K
Reasoning
6/10
Speed
4/10
Fine-tuning
Streaming
Status
Current downloadable open-weight model; no first-party hosted API deployment verified
Input
No official hosted API price; downloadable weights are available
Output
No official hosted API price; downloadable weights are available
View model
→
Continued pretraining, supervised fine-tuning, reinforcement-learning research, language-model research, custom deployment, and applications requiring a transparent open training pipeline.
Type
General Purpose
Context
66K
Reasoning
5/10
Speed
7/10
Fine-tuning
Streaming
Status
Current and downloadable open-weight model
View model
→
Open, self-hosted chat assistants, instruction following, local coding help, tool-calling agents, and model research
Type
General Purpose
Context
66K
Reasoning
7/10
Speed
7/10
Tool use
Fine-tuning
Streaming
Status
Current open-weight model; available for download and self-hosted inference
View model
→
Open research, mathematical reasoning, code generation, logic, long-context experimentation, and local self-hosted inference
Type
Reasoning
Context
66K
Reasoning
9/10
Speed
6/10
Fine-tuning
Streaming
Status
Active open-weight model
View model
→
Open-weight research, continued pretraining, fine-tuning, programming, mathematics, reading comprehension, and long-context language-model experiments
Type
General Purpose
Context
66K
Reasoning
6/10
Speed
4/10
Fine-tuning
Streaming
Status
Current; open-weight and downloadable
Input
No official hosted API price; downloadable weights
Output
No official hosted API price; downloadable weights
View model
→
Open-weight chat, tool-using assistants, multi-turn dialogue, synthetic data generation, self-hosting, and model research
Type
General Purpose
Context
66K
Reasoning
7/10
Speed
5/10
Tool use
Fine-tuning
Streaming
Status
Available open-weight checkpoint; superseded by Olmo 3.1 32B Instruct for newer instruction-tuned deployments
Input
No official first-party hosted API price; downloadable open weights
Output
No official first-party hosted API price; downloadable open weights
View model
→
Open-model reasoning research, mathematics, coding, long-context analysis, and self-hosted deployment
Type
Reasoning
Context
66K
Reasoning
8/10
Speed
5/10
Fine-tuning
Streaming
Status
Available as an open-weight model; superseded by the newer Olmo 3.1 32B Think model
Input
No official hosted API pricing; downloadable weights
Output
No official hosted API pricing; deployment and inference costs are user-managed
View model
→
Math-focused RLVR research, reinforcement-learning experiments, verifiable reward design, open-model evaluation, and further fine-tuning
Type
Reasoning
Context
66K
Reasoning
7/10
Speed
7/10
Fine-tuning
Status
Current open-weight research checkpoint
View model
→
Mathematical reasoning, coding, complex multi-step tasks, open-model research, local deployment, and reinforcement-learning research
Type
Reasoning
Context
66K
Reasoning
8/10
Speed
5/10
Fine-tuning
Streaming
Status
Current open-weight model
View model
→
Open coding-model research, reinforcement learning from verifiable rewards, code-generation experiments, automated evaluation, and fine-tuning
Type
Coding
Context
66K
Reasoning
5/10
Speed
7/10
Fine-tuning
Status
Current, open-weight, experimental RL-Zero coding checkpoint
View model
→
Self-hosted chat assistants, instruction following, tool-use agents, synthetic data, and open-model research
Type
General Purpose
Context
66K
Reasoning
7/10
Speed
4/10
Tool use
Fine-tuning
Status
Current; open-weight and downloadable
View model
→
English audio transcription, captions, meetings, lectures, podcasts, accessibility applications, and open ASR research
Type
Other
Reasoning
1/10
Speed
8/10
Audio input
Streaming
Status
Current open-weight model
View model
→
Open English speech-to-text transcription, captioning, meetings, lectures, calls, podcasts, and local ASR research
Type
Other
Reasoning
1/10
Speed
6/10
Audio input
Streaming
Status
Available open-weight model
Input
No official hosted API pricing; released for open-weight deployment
Output
No official hosted API pricing; released for open-weight deployment
View model
→
Open, self-hosted English transcription for meetings, lectures, calls, podcasts, accessibility, and speech-recognition research
Type
Other
Reasoning
1/10
Speed
5/10
Audio input
Streaming
Status
Current open-weight model; downloadable and usable for local inference
Input
No official hosted API pricing; self-hosted open weights
Output
No official hosted API pricing; self-hosted open weights
View model
→
English short- and long-form speech transcription, meeting and call transcription, lecture captioning, podcast processing, broadcast transcription, and speech analytics
Type
Other
Reasoning
1/10
Speed
6/10
Audio input
Status
Current open-weight model
Input
No official hosted API price; self-hosted checkpoint
Output
No official hosted API price; self-hosted checkpoint
View model
→
Open English speech transcription, meeting and podcast transcription, captioning, timestamped audio indexing, and ASR research
Type
Other
Reasoning
1/10
Speed
7/10
Audio input
Status
Current open-weight model; available for download and self-managed inference
Input
No official hosted API pricing; open weights available for self-hosted use
Output
No official hosted API pricing; output is transcribed text generated during local inference
View model
→
Efficient self-hosted English transcription, speech research, edge-oriented experiments, and applications where a small open ASR checkpoint is preferred.
Type
Lightweight
Reasoning
1/10
Speed
8/10
Audio input
Status
Available open-weight checkpoint
Input
No official hosted API price; self-hosted checkpoint
Output
No official hosted API price; self-hosted checkpoint
View model
→
Open-model research, efficient local text generation, language-model evaluation, and fine-tuning
Type
Lightweight
Context
4K
Reasoning
4/10
Speed
8/10
Fine-tuning
Streaming
Status
Available open-weight release
View model
→
Efficient satellite-image and Earth-observation embeddings, remote-sensing research, and downstream classification or segmentation
Type
Multimodal
Reasoning
1/10
Speed
8/10
Multimodal
Image input
Fine-tuning
Status
Current; downloadable open-weight research model
View model
→
Satellite-image and Earth-observation embeddings, remote-sensing representation learning, geospatial classification, segmentation, and downstream fine-tuning
Type
Multimodal
Reasoning
1/10
Speed
7/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight model
Input
No official hosted API pricing; downloadable weights
Output
No official hosted API pricing; produces feature representations and embeddings
View model
→
Efficient Earth observation embeddings, satellite image and time-series representation learning, geospatial classification, segmentation, and large-scale remote-sensing inference
Type
Multimodal
Reasoning
2/10
Speed
9/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight model
View model
→
Satellite-image embeddings, remote-sensing representation learning, geospatial segmentation, land-cover analysis, and Earth observation research
Type
Multimodal
Reasoning
1/10
Speed
7/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight Earth observation foundation model
Input
No official public hosted API price; downloadable weights are available
Output
No official public hosted API price; model output is primarily embeddings and task representations
View model
→
Open-weight language-model research, long-context text generation, local deployment, continued pretraining, and fine-tuning
Type
General Purpose
Context
66K
Reasoning
6/10
Speed
8/10
Fine-tuning
Streaming
Status
Current open-weight model
View model
→
Self-hosted instruction following, open post-training research, mathematics, coding, evaluation and domain adaptation
Type
General Purpose
Context
131K
Reasoning
7/10
Speed
7/10
Fine-tuning
Status
Available for download; superseded by Llama-3.1-Tulu-3.1-8B
View model
→
Self-hosted instruction following, reasoning, mathematics, coding, open-model research, and reproducible post-training experiments
Type
General Purpose
Context
131K
Reasoning
8/10
Speed
3/10
Fine-tuning
Status
Available open-weight model
Input
No official hosted API pricing; downloadable weights
Output
No official hosted API pricing; downloadable weights
View model
→
Large-scale research, instruction-following evaluation, open-weight post-training research, mathematical reasoning, coding benchmarks, and self-hosted experimentation.
Type
General Purpose
Context
8K
Reasoning
8/10
Speed
2/10
Fine-tuning
Streaming
Status
Available as an open-weight research and educational model; no hosted inference provider deployment is currently listed on its Hugging Face model page.
Input
No official first-party hosted API price; self-hosted open weights
Output
No official first-party hosted API price; self-hosted open weights
View model
→
Multimodal research, vision-language experiments, image generation, visual question answering, dense computer-vision tasks, and academic benchmarking
Type
Multimodal
Reasoning
4/10
Speed
3/10
Multimodal
Image input
Media output
Status
Open-weight research release; publicly available inference code and checkpoints; legacy relative to Ai2's newer multimodal models
View model
→
Multimodal research, image understanding and generation, audio and video understanding, spatial prediction, embodied AI and robotic-manipulation experiments, and self-hosted academic prototyping.
Type
Multimodal
Reasoning
5/10
Speed
2/10
Multimodal
Image input
Audio input
Status
Open-weight research release; publicly accessible checkpoints and source code; no official hosted inference API or commercial token pricing identified.
Input
No official hosted API pricing; self-hosted research checkpoints
Output
No official hosted API pricing; self-hosted research checkpoints
View model
→
Open-vocabulary monocular 3D detection, spatial perception, robotics research, augmented reality, and lifting 2D prompts into metric 3D boxes
Type
Other
Reasoning
2/10
Speed
2/10
Multimodal
Image input
Fine-tuning
Status
Current open-weight research model
Input
No official hosted API pricing; downloadable checkpoint
Output
No official hosted API pricing; downloadable checkpoint
View model
→
Reasoning
0/10
Speed
0/10
View model
→
