Assertion Partly supported AI assessment confidence: 88% certainty 5/5 debate potential 1/5

Madra: Llama 3 70B is over 10x cheaper than GPT-4

Sundeep Madra · Ep8. AI Models, Data Scaling, Enterprise & Personal AI | BG2 with Bill Gurley & Brad Gerstner · Bg2 Pod · Apr 29, 2024 · at 5:16

Sunny Madra (Head of Groq Cloud) breaks down the pricing differential between hosting OpenAI's GPT-4 and running Meta's open-source Llama 3 70B model.

0:00 / 0:29exact quote · 29.9s

▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark

“Price performance, right? You get probably on, from a GPT-IV, you're more than 10 times cheaper. Right. And you're. Yeah. Yeah. 10 times cheaper. And well, let me just tell you, GPT-IV is 10 dollars per million tokens input and 30 dollars per million token output. And Lama three 70 B is 60 cents for a million tokens input and 70 cents per million tokens output.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

Inference on Meta's Llama 3 70B is more than 10 times cheaper than OpenAI's GPT-4, costing $0.60 per million input tokens and $0.70 per million output tokens compared to $10 input and $30 output for GPT-4.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

Sundeep Madra's core assertion that Llama 3 70B inference is over 10x cheaper than OpenAI's GPT-4 was accurate as of April 2024, but his specific quoted price for output tokens ($0.70/M) slightly understated the prevailing hosting rate. As of April 2024, OpenAI's official GPT-4 Turbo API pricing was exactly $10.00 per 1 million input tokens and $30.00 per 1 million output tokens. For Llama 3 70B, prominent fast inference provider Groq offered $0.59 per 1M input tokens (matching Madra's $0.60) and $0.79 per 1M output tokens (compared to Madra's stated $0.70). Across other providers (e.g., DeepInfra at $0.52–$0.59 input / $0.75–$0.79 output, Fireworks/Together at ~$0.90), input costs were ~16.7x–16.9x cheaper and output costs were ~38.0x–42.8x cheaper than GPT-4 Turbo, easily exceeding the 10x threshold.

OpenAI published official pricing for GPT-4 Turbo (gpt-4-turbo / gpt-4-0125-preview) at $10.00 per 1M input tokens and $30.00 per 1M output tokens as of April 2024.
Groq launched Llama 3 70B API inference in April 2024 at $0.59 per 1M input tokens and $0.79 per 1M output tokens.
The actual cost ratio between GPT-4 Turbo and Llama 3 70B on Groq was ~16.9x cheaper for input tokens and ~38.0x cheaper for output tokens, surpassing the 'more than 10 times cheaper' claim.

Checked against

[1] openai.com [2] openrouter.ai [3] galileo.ai [4] portkey.ai [5] adasci.org [6] inferencebench.io [7] koonka.ai [8] llmpricecheck.com

"0.59" "0.79" groq "llama 3" april 2024 openai gpt 4 turbo pricing 10 30 per million tokens april 2024 groq api pricing llama 3 70b april 2024 "together ai" "llama 3 70b" pricing april 2024 llama 3 70b pricing 0.60 0.70 per million tokens "groq" "0.59" "0.79" "llama 3 70b" "openai" "pricing" "april 2024" "gpt-4" "$10" "$30"

assessed as of 2026-09-05

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

Meta GPT-4 Llama 3 70B OpenAI

⚑ Spot an error? Report it

More from Sundeep Madra

Assertion Not checkable as stated

Madra: Developers replace OpenAI with Llama 3 with no performance loss

“And people are doing a direct replacement with open AI across the board. They come to us. So they come to, you know, all the different providers and they replace out open AI. And they don't really see any performance impact or any reasoning impact”

Sundeep Madra Apr 29, 2024 ▶ 4:56 Ep8. AI Models, Data Scaling, Enterprise & Personal AI | BG2 with Bill Gurley & Brad Gerstner · Bg2 Pod

Assertion Contradicted

Madra: 82% of enterprises use or plan to use open-source AI

“82% of the respondents said they are either already on open source or will move to open source.”

Sundeep Madra Apr 29, 2024 ▶ 21:54 Ep8. AI Models, Data Scaling, Enterprise & Personal AI | BG2 with Bill Gurley & Brad Gerstner · Bg2 Pod

Assertion Not checkable as stated

Madra: Llama 3 became Groq's most popular model in 48 hours

“And for us, we saw within the first 48 hours, it become the most power, most popular model that we run on Grok. And so really replacing what? Replacing Mixtrel eight by seven for us, which was, you know, generally considered the best open source model at that …”

Sundeep Madra Apr 29, 2024 ▶ 4:31 Ep8. AI Models, Data Scaling, Enterprise & Personal AI | BG2 with Bill Gurley & Brad Gerstner · Bg2 Pod

Madra: Llama 3 70B is over 10x cheaper than GPT-4

More from Sundeep Madra

Turn any episode into a week of clips.