Models
DeepSeek
Token Throughput per GPU vs. Interactivity

@SemiAnalysis_X · Feb 25, 2026

Token Throughput per GPU vs. Interactivity

@SemiAnalysis_'s scatter plot drew 2.7K views on X: 0.1x @semianalysis_'s median post and #1,048 of 1,162 AI chart posts we hold by views. Read Oct 12, 2026.

Token throughput per GPU vs interactivity for DeepSeek R1

XData Scatter plot DeepSeek R1 All DeepSeek modelsB200 (Dynamo SGLang, MTP)B200 (Dynamo SGLang, TRT)B200 (Dynamo TRT)B200 (Dynamo TRT, MTP)B200 (SGLang)B200 (TRT)B200 (TRT, MTP)B200 (SGLang, MTP)

@SemiAnalysis_: Token Throughput per GPU vs. Interactivity — Published by @SemiAnalysis_ on X, Feb 25, 2026. Source: x.com

11/ 100

ViralChart score

Quiet

Fewer views and a lower engagement rate than the median post.

#1,148 of 1,162 AI chart posts we hold · #24 of 24 @semianalysis_ posts. Score = views 40%, likes 30%, engagement rate 15%, save rate 15%, each as a percentile of every AI chart post we hold.

Our read

It lands in the Quiet quadrant, at #1,148 of 1,162 AI chart posts.

Its 2.7K views are 0.07x this company's 40.3K median, ranking #24 of 24 @SemiAnalysis_ posts.

The score of 11 sits at #1,148 of 1,162 AI chart posts and #24 of 24 @SemiAnalysis_ posts.

Engagement of 0.40% is 0.55x the 0.73% median among AI chart posts, at the 18th percentile.

0.1x@semianalysis_'s median post

2.7K views, where @semianalysis_'s typical post gets 40.3K.

0.0xthe median scatter plot post

Across 68 scatter plot posts we hold, the median gets 73.7K views.

2charts in this one post

The numbers belong to the post, so every chart in it shares them and it is ranked once.

bottom 18%engagement rate

0.40% of viewers liked, reposted, quoted or replied. The median post gets 0.73%.

Public X numbers from our read on Oct 12, 2026. Peers: 1,162 AI chart posts we hold and 24 from @semianalysis_.

Where it sits

Reach against resonance

Each dot is a post. Right is more views, up is a higher engagement rate. The lines are the medians, so the four corners are the four kinds of post.

Among 1,162 AI chart posts we hold

Score #1,148 of 1,162

Views	2.7K	#1,048	bottom 10%
Likes	10	#1,115	bottom 4%
Engagement rate	0.40%	#955	bottom 18%
Save rate	0.036%	#947	bottom 19%

Among 24 @semianalysis_ posts

Score #24 of 24

Views	2.7K	#24	bottom 2%
Likes	10	#22	bottom 10%
Engagement rate	0.40%	#19	bottom 23%
Save rate	0.036%	#20	bottom 19%

Growth

One read so far

Views2.7K

Age at our read230 days

Views per day of life12views divided by days since it was posted

Dots are our reads. The dotted line from the day it was posted to our first read is a gap, not data.

@semianalysis_'s posts over time

Every post from this account we hold, by date and views. This one is marked; the line is the running median of the last 10.

News · Feb 25, 2026

@SemiAnalysis_ scatter plot on token throughput per GPU

A 16-model scatter plot on token throughput per GPU against interactivity drew 2.7K views and a 0.40% engagement rate.

What happened

@SemiAnalysis_ posted a scatter plot on token throughput per GPU vs. interactivity for DeepSeek R1 on February 25, 2026. The post carried 2 charts, the other being TCO per GPU vs Interactivity, and came with 1 other post that day.

The numbers

It reached 2.7K views, #1,048 of 1,162 AI chart posts, against a 39.9K median among AI chart posts. Its score of 11 ranked #1,148 of 1,162 AI chart posts and #24 of 24 @SemiAnalysis_ posts.

The lesson

A 16-model scatter plot drew 2.7K views, 0.07x this company's 40.3K median, and a 0.40% engagement rate against a 0.73% median among AI chart posts. The chart shows token throughput per GPU values from 700 to 1900.

What it means

At #24 of 24 @SemiAnalysis_ posts and #1,148 of 1,162 AI chart posts, this scatter plot sits below the account's own 40.3K median views. The 0.40% engagement rate is 0.55x the 0.73% median among AI chart posts.

Written by ViralChart from public numbers read on Oct 12, 2026. Every number is checked against our data before it is shown.

Updates: what @semianalysis_ posted next

Data and API

19 points read off this chart

Estimated from the image 50%

Group	Value	Read
B200 (SGLang)	1900 tok/s/gpu	printed
B200 (SGLang)	1200 tok/s/gpu	printed
B200 (SGLang)	1000 tok/s/gpu	printed
B200 (SGLang)	900 tok/s/gpu	printed
B200 (SGLang)	800 tok/s/gpu	printed
B200 (SGLang)	700 tok/s/gpu	printed
B200 (SGLang)	600 tok/s/gpu	printed
B200 (SGLang)	400 tok/s/gpu	printed
MI355X (SGLang)	1600 tok/s/gpu	printed
MI355X (SGLang)	1200 tok/s/gpu	printed
MI355X (SGLang)	1000 tok/s/gpu	printed
MI355X (SGLang)	900 tok/s/gpu	printed

Showing 12 of 19. All points, the redraw and the maker

Max plan · API

GET https://viral-chart.com/api/posts/a6fe0284a13e.json
GET https://viral-chart.com/api/charts/c48f0c87065f.json
GET https://viral-chart.com/api/charts/c48f0c87065f.skill.md

Make one like this Open the post JSON Chart data JSON Skill file API docs

Static files in this prototype. The post JSON holds the public numbers, ranks, score, quadrant, our reads and, when checked, our copy.

API · Max

Use this chart via API

On the Max plan you can pull this chart's numbers with the provenance of each one, as JSON or CSV, with its redraw spec, or call it from an agent over MCP. Chart id c48f0c87065f.

See the Max plan API docs

	Path	Returns	Plan
GET	`/api/v1/charts/c48f0c87065f`	Title, publisher, date, form, point count; with a Max key, the data link	Open
GET	`/api/v1/charts/c48f0c87065f/data`	Every point with its provenance and confidence, plus the redraw spec. Add ?format=csv for CSV	Max
GET	`/api/v1/charts/c48f0c87065f/render?view=1`	The draw-only view this page uses	Open
POST	`/mcp`	Tool get_chart_data with this id, for Claude, Cursor and other agents	Max

curlMax

curl -s https://viral-chart.com/api/v1/charts/c48f0c87065f/data \
  -H "Authorization: Bearer $VIRALCHART_KEY"

Claude CodeMax

claude mcp add --transport http viralchart https://viral-chart.com/mcp \
  --header "Authorization: Bearer $VIRALCHART_KEY"

60 requests a minute per key; over that you get 429 with Retry-After. Up to 5 keys per account, made on your account page and shown once. Without Max, the data route answers 402 with a link to the plans. The numbers describe a chart its publisher owns, and every file carries a license note.

More charts about DeepSeek R1

DeepSeek R1: AI Model Performance Comparison Across Benchmarks

OfficialX · DeepSeekJan 20, 2025

AI Model Performance Comparison Across Benchmarks12.6M views · 35.1K likes