family member
Sarvam logo

Sarvam-1

111 tokens a second on Linux · NVIDIA 15 GB · 4 CPUs · 31 GB RAM. No hosted API sells it, so this network is the place to run it.

2 builds of Sarvam-1, 7 signed runs on real machines: different packagers, precisions and engines, each a different file with its own numbers. 2B parameters, as the vendor states.

111tok/s, fastest machine
$1.22per M tokens · no API sells it
8Kcontext proved by recall
3.2 GBmemory, smallest build
6providers measured it
Use it in your own room Ask Evo in your gramx and it runs on a machine that proved it, at the metered $1.22 per million tokens. On your own machines it costs you nothing.Measure it on your machine A gram works through the catalog by itself and signs every result. 6 providers have measured Sarvam-1; every run is a point on the providers board.Compare every build Put it beside every other model in one table: speed, price, context and the API price, one row each.
best on ollama · 1 build

proved 8K context, by bartowski Q4_K_M

178.2 tok/s peak, by bartowski Q4_K_M

best on vllm · 1 build

no context proved on this engine yet

Speed against price, every machine

Up and left is faster and cheaper · the line joins machines no other beats on both · price on a log scale

$1$10$100USD per million tokenstok/sLinux · NVIDIA 15 GB · 4 CPUs · 31 GB RAM · 4-bit by bartowski · 111.22 tok/s · $1.2208/MMac · Apple silicon · 10 CPUs · 16 GB RAM · 4-bit by bartowski · 62.17 tok/s · $3.1624/MWindows · NVIDIA 4 GB · 8 CPUs · 8 GB RAM · 4-bit by bartowski · 23.48 tok/s · $7.5107/MWindows · Intel 2 GB · 8 CPUs · 16 GB RAM · 4-bit by bartowski · 12.97 tok/s · $5.563/MWindows · NVIDIA 12 GB · 16 CPUs · 16 GB RAM · 4-bit by bartowski · 2.01 tok/s · $127.936/M

Same build, different machines

Q4_K_M by bartowski on 6 machines: the fastest decodes 39× as fast as the slowest. One dot per machine, tokens per second.

  1. NVIDIA GeForce RTX 3080 Ti×1178 tok/s
  2. Tesla T4×1111 tok/s
  3. Apple M4×162 tok/s
  4. NVIDIA GeForce GTX 1650×123 tok/s
  5. Intel(R) UHD Graphics 730×113 tok/s
  6. AMD Radeon(TM) Vega 8 Mobile Graphics×14.5 tok/s

What it fits on

The smallest build needs about 3.2 GB; 13 of 13 kinds of machine on the network can hold it. Bars: each kind's memory (its card's, else RAM) · ticks: every build, at the size it held plus a fifth for the prompt.

  1. NVIDIA GeForce RTX 3050 Laptop GPU4 GB on the card1 of 2
  2. NVIDIA GeForce GTX 16504 GB on the card1 of 2
  3. AMD Radeon(TM) Vega 8 Mobile Graphics7 GB RAMevery build
  4. Windows, CPU only · 12 CPUs7 GB RAMevery build
  5. Apple silicon8 GB RAMevery build
  6. Apple M18 GB RAMevery build
  7. NVIDIA GeForce RTX 3080 Ti12 GB on the cardevery build
  8. Tesla T415 GB on the cardevery build
  9. Intel(R) UHD Graphics 73016 GB RAMevery build
  10. Intel(R) Iris(R) Xe Graphics16 GB RAMevery build
  11. Apple M416 GB RAMevery build
  12. 2× Tesla T429 GB on the cardevery build
  13. Linux, CPU only · 4 CPUs31 GB RAMevery build

How far back it remembers, machine by machine

Filled: the needle came back at that prompt size · outlined: asked and lost · blank: not asked · trained for 8K. 2 walks ended because the model forgot, 0 because the machine could not hold the prompt.

machine · build4K8K16K32K64K128Kwhy it stopped
Gram ·12954dNVIDIA GeForce GTX 1650 · Q4_K_M by bartowski8K
Gram ·a5be46Tesla T4 · Q4_K_M by bartowski4K the model lost the needle at 8K
Kunal HindochaNVIDIA GeForce RTX 3080 Ti · Q4_K_M by bartowski—
nitin singhApple M4 · Q4_K_M by bartowski—
Amit KhetanIntel(R) UHD Graphics 730 · Q4_K_M by bartowski—
Gram ·a5be46Tesla T4 · BF16 by sarvamai— the model lost the needle at 4K

Carbon per million tokens

Lowest build on each machine · 1 of 1 read off a power sensor, the rest modelled from the parts · fewer grams is better

  1. Gram ·a5be46251 g1.91 J/tok · sensor

Carbon is location-based: signed energy × the yearly average of the grid where the machine signed its place (world grid average (place not signed); Ember Global Electricity Review 2025 (2024 generation), location-based). Facility overhead is not included. A machine on its own solar is not credited until it can prove it.

Measured by

The people whose machines signed these numbers

proved by at least one build: Structured output · Sustained output · Follows instructions

Every build

Each its own file with its own page: the machines that ran it and their signatures.

buildenginewhose workmemorycontext provedbest tok/smachines
Q4_K_M by bartowskihf.co/bartowski/sarvam-1-GGUF:Q4_K_Mollamaquantizer · bartowski3.2 GB8K178.26
BF16 by sarvamaisarvamai/sarvam-1vllmvendor · sarvamai5.7 GB——1