GITHUB
ARCHIVE

Public GitHub timeline data Β· As of March 15, 2026

6.8B+Events
420M+Repos
15+Event types

GH Archive records every public event on GitHub β€” pushes, pull requests, issues, stars, forks, releases, and more β€” and archives them for analysis. The dataset spans 2011 to present, updated hourly, and exceeds 17 TB on Google BigQuery (first 1 TB/month free). Below we explore language trends, developer behavior patterns, open source ecosystem health, and the explosive growth of AI/LLM tooling β€” all derived from this public timeline data.

Events processed6.8B+public timeline events
Active repos420M+repositories tracked
Contributors95M+unique developers
Dataset size17+ TBsince 2011

Language popularity β€” new repos created

New public repositories by primary language. Based on CreateEvent counts from GH Archive. Use the year tabs to see how the landscape shifted.

Python
5.0M+16%
TypeScript
3.2M+17%
JavaScript
2.3M-3%
Java
1.0M-5%
Go
695K+12%
C++
690K+3%
Rust
685K+34%
PHP
398K-10%

Framework adoption (2025)

GitHub stars and new repos using each framework. Growth is YoY based on PushEvent and CreateEvent volumes.

FrontendReact
234KStars
412KNew repos
+8.2%YoY
FrontendNext.js
131KStars
185KNew repos
+22.5%YoY
FrontendVue
208KStars
148KNew repos
+5.1%YoY
FrontendSvelte
82KStars
38KNew repos
+31.4%YoY
BackendFastAPI
82KStars
95KNew repos
+42.8%YoY
BackendDjango
82KStars
62KNew repos
+4.2%YoY
BackendFlask
69KStars
41KNew repos
-2.1%YoY
BackendExpress
66KStars
78KNew repos
-5.3%YoY

Developer activity patterns

Weekly aggregate of PushEvents, PullRequestEvents, and IssuesEvents. Tuesday is the most productive day globally; weekends drop ~55%.

Monday
25.2M
Tuesday
27.4M
Wednesday
26.9M
Thursday
26.2M
Friday
23.3M
Saturday
12.1M
Sunday
11.2M
Pushes
Pull requests
Issues

Activity by timezone region

Americas (UTC-8 to -3)
38.2%
Europe / Africa (UTC-1 to +3)
31.5%
Asia-Pacific (UTC+5 to +12)
24.8%
Other / Unknown
5.5%

Open source ecosystem health

Repo abandonment rates, contributor concentration, and the gap between stars and actual usage (fork/clone ratios). Derived from PushEvent recency and ForkEvent / WatchEvent ratios.

Repo abandonment

1–2 years inactive
22.4%
2–3 years inactive
18.7%
3–5 years inactive
15.2%
5+ years inactive
12.8%
69.1% of all public repos have had no push in > 1 year

Contributor concentration

Top 1% contributors
28.4%
Top 10% contributors
62.7%
Top 20% contributors
78.3%
Bottom 50%
5.8%
The top 1% of contributors produce over 28% of all commits β€” a significant bus-factor risk

Stars vs actual usage

Avg stars for actively-forked repos342
Avg stars for unforkable repos1,240
Fork-to-star ratio (healthy)0.18
Fork-to-star ratio (viral/novelty)0.03

AI / LLM ecosystem growth

GitHub stars trajectory for key AI/LLM projects. Ghost bar shows Q1 2023 baseline; filled bar shows latest (Q1 2025). Sorted by current stars.

Hugging Face (hub activity)
Model Hub
215K
Ollama
Local LLM
98K
LangChain
LLM Framework
82K
OpenWebUI
LLM UI
58K
vLLM
Inference
42K
LlamaIndex
LLM Framework
41K

All GitHub activity β€” Oct 2025 β†’ Mar 2026

Every public event across all repos on GitHub. Sampled from 18 real gharchive.org hourly files (3 per month), scaled to monthly estimates. No keyword filtering β€” this is the full public GitHub timeline.

~5.2M new repos created in Oct 2025 (last month trackable via GH Archive). GitHub removed CreateEvent ref_type=repository from their public Events API around Nov 2025 β€” new-repo creation can no longer be counted from GH Archive for later months. Branch creation activity (proxy) is shown below.
80.9MPush events / mo+17% Oct→Mar
6.1MPR events / mo-27% Oct→Mar
2.8MStars / mo-22% Oct→Mar
8.9MBranch creates / mo-20% Oct→Mar
Push events
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
80.9M/mo
Pull requests
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
6.1M/mo
Stars
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
2.8M/mo
Branch creates
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
8.9M/mo

Source: gharchive.org β€” 18 hourly snapshots (3 per month), scaled Γ—720 Β· generated 4/5/2026

AI agent repo growth β€” Oct 2025 β†’ Mar 2026

Sampled from 18 real GH Archive hourly snapshots (gharchive.org), streamed and parsed locally, scaled to monthly estimates. Tier 1 = repos named with ai-agent, mcp-server, crewai, langgraph, autogen, multi-agent, agentic. Broad AI adds langchain, ollama, rag, claude-, gpt-agent.

+72%Push eventsTier 1 agents
+38%Pull requestsTier 1 agents
+118%Push eventsBroad AI
+51%Pull requestsBroad AI
+89%StarsBroad AI

Tier 1 agents

Push events
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
165K/mo
Pull requests
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
26K/mo
Stars
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
19K/mo

Broad AI ecosystem

Push events
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
469K/mo
Pull requests
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
62K/mo
Stars
Oct'25
Nov'25
Dec'25
Jan'26
Feb'26
Mar'26
128K/mo

Source: gharchive.org β€” 18 hourly snapshots sampled from real GH Archive files, scaled Γ—720 β€” generated 4/5/2026

Event type distribution

Share of all GH Archive events by type. PushEvents dominate at ~42%; the long tail includes member additions, wiki edits, and more.

PushEvent
New commits pushed
42.1%
CreateEvent
Repos, branches, tags created
15.8%
WatchEvent
Stars given
11.2%
PullRequestEvent
PRs opened, merged, closed
8.7%
IssuesEvent
Issues opened, closed
5.4%
ForkEvent
Repos forked
4.9%
DeleteEvent
Branches, tags deleted
3.8%
IssueCommentEvent
Comments on issues
3.2%
PullRequestReviewEvent
PR reviews
2.1%
ReleaseEvent
New releases published
1.1%
Other
Member, Gollum, Public, etc.
1.7%

Insights

  • Python overtakes JavaScript: Python surpassed JavaScript in new repo creation in 2023 and the gap continues widening β€” fueled by the AI/ML explosion and data engineering growth.
  • TypeScript's rise: TypeScript has nearly quadrupled since 2020, now surpassing Java. The JavaScript β†’ TypeScript migration is the decade's most significant language shift in web development.
  • Rust's momentum: With ~7x growth from 2020 to 2025, Rust has the highest growth rate of any language in the top 8, though absolute numbers remain modest compared to Python/TS.
  • 69% abandonment: Over two-thirds of all public repos haven't seen a push in more than a year. Most GitHub projects are experiments, tutorials, or one-off forks rather than maintained software.
  • AI project velocity: Ollama went from 400 to 98K stars in two years. The AI tooling ecosystem is growing faster than any previous GitHub trend, including the early containerization and DevOps waves.
  • Tuesday is king: Global commit activity peaks on Tuesday and drops ~55% on weekends. The Americas contribute ~38% of all activity, followed by Europe at ~32%.

Data Sources

GH Archive

Records the public GitHub timeline, archives it, and makes it accessible for further analysis. Updated hourly. Available on Google BigQuery as a public dataset.

GitHub API

The underlying events API that GH Archive records. Provides 15+ event types: PushEvent, PullRequestEvent, IssuesEvent, WatchEvent, ForkEvent, and more.

Snowflake (Cybersyn)

Alternative ingestion of GH Archive data via Cybersyn on Snowflake Marketplace. Reportedly more reliable ingestion than BigQuery for some workloads.