Boarding pass · Best value
claude-3.5-sonnet
claude-3.5-sonnet · Anthropic
- Gate
- Agentic coding
- Ratings
- 7
- Recommend
- 100%
- Value · Quality · Speed
- 4.6 · 4.7 · 3.7
Value
4.6
out of 5 for value
CS-2026-01
Now boarding · category leaders
Top of the registry
Highest average in each score across every model with 5+ verified ratings. Ties go to the model with more ratings.
Boarding pass · Best quality
claude-3.5-sonnet
claude-3.5-sonnet · Anthropic
- Gate
- Agentic coding
- Ratings
- 7
- Recommend
- 100%
- Value · Quality · Speed
- 4.6 · 4.7 · 3.7
Quality
4.7
out of 5 for quality
CS-2026-01
Boarding pass · Fastest
Claude 3 Haiku
claude-3-haiku · Anthropic
- Gate
- Autocomplete
- Ratings
- 5
- Recommend
- 80%
- Value · Quality · Speed
- 4.4 · 3.2 · 5.0
Speed
5.0
out of 5 for speed
CS-2026-02
As of September 23, 2026, claude-3.5-sonnet leads the 2026 DevPass Model Census on value for money (4.6/5 across 7 verified ratings), claude-3.5-sonnet leads on output quality (4.7/5), and Claude 3 Haiku leads on speed (5.0/5). 7 DevPass developers have filed 18 verified ratings across 3 models. The most-rated model is claude-3.5-sonnet with 7 ratings and a 100% recommend rate.
Departures · full registry
Every model on the registry
Click a column to sort, or narrow the board by vendor, use case, and rating count. Filters live in the URL, so a view can be shared.
Showing 3 of 3 models · sorted by value descendingDOC. CS-2026
Departures · Coding models
Rank = registry position by value · refreshed every 5 min
| Registry rank | Status | ||||||
|---|---|---|---|---|---|---|---|
| 1 | claude-3.5-sonnet Anthropic · claude-3.5-sonnet · mostly agentic coding | 4.6 out of 5 value | 4.7 out of 5 quality | 3.7 out of 5 speed | 100% | 7 | Cleared: 90% or more of raters would recommend this model |
| 2 | Claude 3 Haiku Anthropic · claude-3-haiku · mostly autocomplete | 4.4 out of 5 value | 3.2 out of 5 quality | 5.0 out of 5 speed | 80% | 5 | Boarding: 75–89% of raters would recommend this model |
| 3 | GPT-4o OpenAI · gpt-4o · mostly agentic coding | 3.7 out of 5 value | 4.2 out of 5 quality | 4.3 out of 5 speed | 83% | 6 | Boarding: 75–89% of raters would recommend this model |
Scores are 1–5 averages. Models need 5+ verified ratings to be listed; totals refresh every few minutes. Last updated September 23, 2026 (UTC).
Conditions of entry
The rules of the registry
Usage-verified, or it doesn't count
You can only rate a model your DevPass workspace has hit with 50+ requests in the past 30 days. Nobody rates a model they read a thread about.
Members only
Every respondent has an active, paid DevPass plan. These are verdicts from people spending their own credits.
No small-sample noise
A model is published only after 5 or more developers rate it, and only aggregates ever leave the building.
One reward per member per quarter
The census runs in quarterly waves. Your first entry of each wave earns a free Reset Pass — rate as many models as you use, but nobody can farm passes.
Reading the census
Model Census FAQ
- How are models scored in the DevPass Model Census?
- Each entry rates one model from 1 to 5 on value for money, output quality, and speed, plus a yes-or-no “would you recommend it”. The registry shows the per-model averages and the share of raters who would recommend the model.
- Who can rate a model?
- Only paid DevPass members, and only for models their workspace has sent at least 50 requests to in the past 30 days. Ratings are tied to verified usage, not opinions from a thread.
- How is the registry ranked?
- Registry rank is the model's position by average value score, with ties broken by number of ratings. The sort and filter controls change what you see on the board but never the underlying rank.
- Why is a model missing from the registry?
- A model appears once 5 or more developers have filed a verified rating on it in 2026. Per-use-case breakdowns are additionally hidden below two entries so no single respondent can be identified.
- How often does the census update?
- Members file entries in quarterly waves and the yearly registry aggregates every wave. The published totals refresh every five minutes.
Your entryWave Q3 · 2026
Rate the models you ship with, stamp a free Reset Pass
Takes two minutes. Your first entry of the wave earns a Reset Pass on your tier, and every entry sharpens the registry for the next developer at the desk.