AI Model Release Dashboard

Google

Gemini 3.1 Flash Image

February 25, 2026

Faster image generation with 4K resolution, improved text rendering, and multi-language support. #1 on Artificial Analysis Image Arena at launch.

Google

Gemini 3 Deep Think

December 3, 2025

Extended-thinking version of Gemini 3 Pro. Rolled out to Ultra subscribers. Designed for the hardest reasoning tasks.

Mistral

Mistral Small 3.2

June 19, 2025

24B parameter open-weight model (Apache 2.0). Improved instruction following and function calling over v3.1.

Mistral

Magistral Small

June 9, 2025

Mistral's first reasoning model with chain-of-thought capabilities. 24B parameters, open-source under Apache 2.0.

Anthropic

Claude Sonnet 4.5

September 28, 2025

Best coding model at launch with 77.2% on SWE-bench Verified. 30+ hour autonomous task horizon.

Anthropic

Claude Opus 4.1

August 4, 2025

Drop-in replacement for Opus 4 with superior performance. Handles complex multi-step agentic tasks with more rigor.

OpenAI

GPT-5.1

November 11, 2025

Warmer personality, improved instruction-following. Adds shopping research and 8 personality options. Retired March 2026.

OpenAI

GPT-5.2

December 10, 2025

Next upgrade to the GPT-5 series. Smarter and more useful for work and learning. 400K context, 128K output.

OpenAI

GPT-5.2-Codex

December 17, 2025

Agentic coding model for long-horizon task completion. Context compaction for multi-window workflows. 56.4% on SWE-Bench Pro.

OpenAI

GPT-5.3-Codex

February 4, 2026

Most capable agentic coding model to date. 25% faster than GPT-5.2-Codex. State-of-the-art on SWE-Bench Pro and Terminal-Bench.

OpenAI

GPT-5.3 Instant

March 2, 2026

Smoother, more helpful everyday conversations. 400K context. 26.8% fewer hallucinations than GPT-5.2 with web search.

OpenAI

GPT-5.4 nano

March 16, 2026

Smallest and cheapest OpenAI model. Optimized for classification, data extraction, ranking, and coding subagents.

OpenAI

GPT-5.4 mini

March 16, 2026

Most capable small OpenAI model. 2x faster than GPT-5.4. Strong on coding, reasoning, and computer use. Free and Go users.

OpenAI

GPT-5.4

March 4, 2026

Most capable frontier model for professional work. Native computer use, 1M context, consolidated coding+reasoning. 33% fewer hallucinations vs GPT-5.2.

Mistral

Mistral Small 4

March 16, 2026

Unified model combining text, vision, reasoning, and coding. MoE 119B total / 6B active params. 256K context. Apache 2.0.

Google

Gemini 3.1 Flash Lite

March 2, 2026

Ultra-efficient Gemini 3.1 model for developers. Fast and cost-effective for high-volume API workloads.

Anthropic

Claude Sonnet 4.6

February 16, 2026

Default model for free and Pro users. 1M token context (beta). Improved computer use, coding, and large-scale data processing.

Google

Gemini 3.1 Pro

February 18, 2026

Advanced reasoning over complex problems. 77.1% on ARC-AGI-2. Available via Gemini API, Vertex AI, and NotebookLM.

Anthropic

Claude Opus 4.6

February 4, 2026

1M token context window. 14.5-hour task completion horizon. Top Finance Agent benchmark. Stronger coding and planning.

Google

Gemini 3 Flash

December 16, 2025

Efficient Gemini 3 model. Includes Gemini 3 Pro reasoning in a faster, cheaper package. Default model in the Gemini app.

Mistral

Ministral 3 (3B/8B/14B)

December 1, 2025

Family of dense small models in 3 sizes × 3 variants (Base/Instruct/Reasoning). Runs on 4GB VRAM. Apache 2.0 license.

Mistral

Mistral Large 3

December 1, 2025

MoE architecture: 41B active / 675B total parameters. Multimodal, 256K context. Apache 2.0. #2 on LMArena OSS leaderboard.

Google

Gemini 3 Pro

November 17, 2025

Best multimodal understanding model. Most powerful agentic and coding model from Google. Replaces Gemini 2.5 Pro.

Anthropic

Claude Opus 4.5

November 23, 2025

Most robustly aligned Claude model. Reclaimed coding crown with 80.9% SWE-bench Verified. Significantly reduced pricing.

Anthropic

Claude Haiku 4.5

October 14, 2025

Fast and affordable Claude 4 family model. Targets latency-sensitive applications. Available on Claude.ai and mobile.

Google

Gemini 2.5 Flash Image

August 25, 2025

Dedicated image generation model from Google. Better instruction following and text rendering than previous image models.

Google

Gemini 2.5 Flash-Lite

June 16, 2025

Ultra-efficient Gemini 2.5 variant optimized for speed and cost. Lowest latency in the Gemini 2.5 family.

OpenAI

o3-pro

June 9, 2025

Extended-thinking version of o3. Thinks longer for more reliable responses on the hardest tasks. ChatGPT Pro and API.

OpenAI

GPT-4.1 mini

April 13, 2025

Fast, capable small model. Replaces GPT-4o mini for paid users. Significant improvements in coding and instruction-following.

OpenAI

GPT-4.1

April 13, 2025

Specialized for coding and instruction-following. Stronger than GPT-4o at precise web dev tasks. API-only release.

Google

Gemini 2.5 Flash

June 16, 2025

Efficient version of Gemini 2.5 (GA). Controllable thinking budget. Best price-performance in Gemini family.

OpenAI

GPT-5

August 6, 2025

Unifies reasoning and standard model capabilities. Automatically switches between fast and deep thinking modes based on task needs.

Anthropic

Claude Opus 4

July 8, 2025

Most intelligent Claude model. Designed for complex, long-horizon tasks and deep reasoning.

Anthropic

Claude Sonnet 4

July 8, 2025

Next generation Claude model. Enhanced coding, reasoning, and agentic task performance.

Meta

Llama 4

April 4, 2025

MoE architecture models: Scout (17Bx16E) and Maverick (17Bx128E). Native multimodal, 10M context window.

OpenAI

o4-mini

April 15, 2025

Cost-efficient reasoning model with vision. Faster than o3, strong coding performance. Tool use support.

OpenAI

o3

April 15, 2025

Flagship reasoning model. Top ARC-AGI scores. Available in API with high/medium/low reasoning effort settings.

Mistral

Mistral Small 3.1

March 16, 2025

Updated small model with vision capabilities added. 24B parameters, 128K context. Apache 2.0 license.

Google

Gemini 2.5 Pro (Preview)

March 24, 2025

State-of-the-art reasoning model. 1M context, native audio/video/image understanding. Top LMArena scores.

OpenAI

GPT-4.5

February 26, 2025

Largest and most capable GPT model. Improved emotional intelligence and world knowledge. 128K context.

Google

Gemini 2.0 Pro (Experimental)

February 4, 2025

Most capable Gemini 2.0 model for complex tasks. 2M context window. Code execution and Google Search grounding.

Google

Gemini 2.0 Flash

February 4, 2025

Generally available release. Upgraded image generation, improved multilingual audio. 1M context. Free tier available.

Mistral

Mistral Saba

February 16, 2025

Specialized for Arabic and South Asian languages. Strong Middle East/North Africa cultural context understanding.

Anthropic

Claude 3.7 Sonnet

February 23, 2025

First hybrid reasoning model. Switchable extended thinking mode. Top scores on SWE-bench Verified (70.3%).

OpenAI

o3-mini

January 30, 2025

Cost-efficient small reasoning model. Three effort levels (low/medium/high). Outperforms o1 on AIME and SWE-bench.

Mistral

Mistral Small 3

January 14, 2025

24B parameter model with 81% MMLU. Apache 2.0 license. Best-in-class small model for latency-sensitive tasks.

Mistral

Codestral 25.01

January 12, 2025

Updated code generation model. 256K context window, supports 80+ programming languages. Available via API.

Meta

Llama 3.3 70B

December 5, 2024

Updated 70B model with performance matching Llama 3.1 405B. More compute-efficient. Open weights.

Google

Gemini 2.0 Flash (Experimental)

December 10, 2024

First Gemini 2.0 model. Multimodal I/O, native tool use, 1M context. Faster than 1.5 Pro. Agentic capabilities.

OpenAI

o1

December 4, 2024

Full reasoning model release. Includes vision support, 200K context. Available in ChatGPT and API.

Anthropic

Claude 3.5 Haiku

November 3, 2024

Fastest Claude model. Matches Claude 3 Opus on many benchmarks. 200K context. API and Bedrock availability.

Mistral

Pixtral Large

November 6, 2024

124B parameter multimodal model. Strong document understanding, charts, and scientific diagrams.

Anthropic

Claude 3.5 Sonnet (v2)

October 21, 2024

Upgraded Sonnet with improved coding (SWE-bench 49%), vision, and computer use capability (beta).

Google

Gemini 1.5 Flash-8B

October 2, 2024

Smallest Gemini model for high-frequency, low-latency tasks. 1M token context. Generally available via API.

Meta

Llama 3.2

September 24, 2024

Multimodal models (11B, 90B) with vision. Lightweight text models (1B, 3B) for edge deployment. Open weights.

OpenAI

o1-mini

September 11, 2024

Smaller, faster reasoning model. Cost-effective for coding and STEM tasks without broad world knowledge needs.

OpenAI

o1-preview

September 11, 2024

Early access reasoning model with extended chain-of-thought. Designed for complex science, coding, and math tasks.

Mistral

Pixtral 12B

September 10, 2024

First multimodal model from Mistral. 12B parameters with vision capabilities. Apache 2.0 license.

Mistral

Mistral Large 2

July 23, 2024

128K context, strong coding and multilingual capabilities. 123B parameters. Apache 2.0 license.

Meta

Llama 3.1

July 22, 2024

Open-weights models: 8B, 70B, 405B. 128K context window. First open model competitive with frontier closed models.

OpenAI

GPT-4o mini

July 17, 2024

Affordable, fast frontier model. 128K context, supports vision. Lower cost alternative to GPT-4o.