Model Release Timeline
Chronological history of frontier LLM releases from the 2017 Transformer paper through the Claude 4.x family in April 2026 — essential context for calibrating what constitutes progress vs expected trajectory.
The frontier LLM release history. Understanding the trajectory helps calibrate what's impressive vs expected.
2017–2019: Foundations
| Date | Model | Lab | Significance |
|---|---|---|---|
| Jun 2017 | Transformer | Architecture that replaced RNNs | |
| Oct 2018 | BERT | Bidirectional transformers; NLP benchmarks shattered | |
| Feb 2019 | GPT-2 | OpenAI | 1.5B params; "too dangerous to release" (briefly) |
2020–2021: Scale
| Date | Model | Lab | Significance |
|---|---|---|---|
| May 2020 | GPT-3 | OpenAI | 175B; in-context learning; API release |
| Jun 2021 | GitHub Copilot | GitHub/OpenAI | First mass-market AI coding tool |
| Aug 2021 | Codex | OpenAI | Code-focused GPT; powers Copilot |
2022: Alignment and ChatGPT
| Date | Model | Lab | Significance |
|---|---|---|---|
| Apr 2022 | PaLM 540B | Scaling laws confirmed at 540B params | |
| Dec 2022 | ChatGPT | OpenAI | Consumer breakout; 1M users in 5 days |
| Dec 2022 | Constitutional AI paper | Anthropic | CAI framework published |
2023: Open Source and Multimodal
| Date | Model | Lab | Significance |
|---|---|---|---|
| Feb 2023 | LLaMA 1 | Meta | Open weights leak; open-source ecosystem born |
| Mar 2023 | GPT-4 | OpenAI | Vision + text; SWE-bench debut model |
| Mar 2023 | Claude 1 | Anthropic | First public Claude; Constitutional AI |
| Jul 2023 | Llama 2 | Meta | Officially open; commercial license |
| Nov 2023 | Mistral 7B | Mistral | Best 7B model; MoE architecture introduced |
| Nov 2023 | Mixtral 8x7B | Mistral | MoE outperforms GPT-3.5; open weights |
| Dec 2023 | Gemini 1.0 | Native multimodal; Ultra/Pro/Nano tiers |
2024: Reasoning and Agents
| Date | Model | Lab | Significance |
|---|---|---|---|
| Mar 2024 | Claude 3 (Haiku/Sonnet/Opus) | Anthropic | Haiku/Sonnet/Opus tiers; Opus #1 on Arena |
| Apr 2024 | Llama 3 8B/70B | Meta | Best open model at each tier |
| May 2024 | GPT-4o | OpenAI | Unified audio/vision/text; realtime API |
| Jun 2024 | Claude 3.5 Sonnet | Anthropic | Reclaimed #1; 50% SWE-bench; Claude Artifacts |
| Sep 2024 | o1 | OpenAI | Reasoning model with internal chain-of-thought |
| Sep 2024 | Llama 3.1 405B | Meta | First open model competitive at frontier |
| Oct 2024 | Claude 3.5 Haiku | Anthropic | Sonnet-level quality at Haiku price |
| Nov 2024 | Qwen 2.5 | Alibaba | Best Chinese open model family |
2025: Agents Take Over
| Date | Model | Lab | Significance |
|---|---|---|---|
| Jan 2025 | DeepSeek V3/R1 | DeepSeek | o1-level reasoning, open weights, 96% cheaper API |
| Feb 2025 | Claude 3.7 Sonnet | Anthropic | Extended thinking; 70.3% SWE-bench Verified |
| Mar 2025 | Gemini 2.5 Pro | Strong reasoning; 1M context becomes standard | |
| Apr 2025 | Claude 3.5 Sonnet refresh | Anthropic | Agent capabilities; Claude Code ships |
| Jun 2025 | GPT-4.5 | OpenAI | Improved instruction following |
| Jul 2025 | Llama 3.1 405B Instruct | Meta | Best open model for coding |
| Oct 2025 | Claude Haiku 4.5 | Anthropic | 73.3% SWE-bench at lowest price tier |
2026 (April): Claude 4 Family
| Date | Model | Lab | Significance |
|---|---|---|---|
| Feb 2026 | Claude Opus 4.6 | Anthropic | 80.8% SWE-bench; 91.3% GPQA |
| Feb 2026 | Claude Sonnet 4.6 | Anthropic | 79.6% SWE-bench; $3/$15 per M tokens |
| Apr 2026 | Claude Opus 4.7 | Anthropic | Latest Opus; released April 16 2026 |
| Apr 2026 | Gemini 3 (unconfirmed) | [unverified] |
[Source: Perplexity research, 2026-04-29 — post-August 2025 dates are from research]
Key Themes by Year
2017–2020: Architecture innovation (Transformer → BERT → GPT-3)
2021–2022: Alignment research + consumer products (ChatGPT)
2023: Open source democratisation (Llama, Mistral); multimodal goes mainstream
2024: Reasoning models (o1); agents start shipping
2025: Agents are the product; open models reach frontier; Claude Code ships
2026: Claude Opus 4.x takes SWE-bench lead; Anthropic overtakes OpenAI on revenue
Key Facts
- GPT-3 (May 2020): 175B parameters; first mass in-context learning via API
- ChatGPT hit 1M users in 5 days (December 2022)
- LLaMA 1 leaked in Feb 2023; open-source ecosystem formed around it
- Claude 3.7 Sonnet (Feb 2025): extended thinking; 70.3% SWE-bench Verified
- DeepSeek R1 (Jan 2025): o1-level reasoning; 96% cheaper; MIT license; GRPO-only training
- Claude Opus 4.6 (Feb 2026): 80.8% SWE-bench Verified; 91.3% GPQA Diamond
- 2026 theme: Anthropic overtakes OpenAI on revenue; Claude Code ships as agentic tool
Connections
- landscape/ai-labs — the companies behind each model release
- llms/claude — Claude family in depth with architecture and capability details
- llms/model-families — current model families and benchmark comparisons
- evals/benchmarks — what SWE-bench, GPQA, and MMLU actually measure
Open Questions
- What does GPT-5 look like and when does it ship?
- Does the open-source tier continue to narrow the gap with frontier models at each generation?
- When does the Transformer architecture get meaningfully supplanted in production?
Related reading