family member
Sarvam logo

Sarvam-1

121 tokens a second on Linux · L4 23 GB · 12 CPUs · 53 GB RAM. No hosted API sells it, so this network is the place to run it.

2 builds of Sarvam-1, 9 signed runs on real machines: different packagers, precisions and engines, each a different file with its own numbers. 2B parameters, as the vendor states.

178tok/s, fastest machine
$1.22per M tokens · no API sells it
8Kcontext proved by recall
5.7 GBmemory, smallest build
8providers measured it
Run it on your gram Install a gram with one command. It works through the catalog on your own machine and signs every result. 8 providers have measured Sarvam-1 so far.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.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

trained 8K · 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$10USD per million tokenstok/sLinux · L4 23 GB · 12 CPUs · 53 GB RAM · 4-bit by bartowski · 121.44 tok/s · $1.7861/MLinux · 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 · GeForce RTX 3080 Ti 12 GB · 16 CPUs · 16 GB RAM · 4-bit by bartowski · 19.76 tok/s · $13.0304/MWindows · Intel(R) UHD Graphics 730 2 GB · 8 CPUs · 16 GB RAM · 4-bit by bartowski · 8.18 tok/s · $8.8254/MWindows · Intel(R) Iris(R) Xe Graphics 1 GB · 8 CPUs · 16 GB RAM · 4-bit by bartowski · 4.83 tok/s · $14.9543/M

Same build, different machines

Q4_K_M by bartowski on 8 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. NVIDIA L4×1121 tok/s
  3. 2× Tesla T4×1111 tok/s
  4. Apple M4×162 tok/s
  5. NVIDIA GeForce GTX 1650×123 tok/s
  6. Intel(R) UHD Graphics 730×113 tok/s
  7. Intel(R) Iris(R) Xe Graphics×16.4 tok/s
  8. AMD Radeon(TM) Vega 8 Mobile Graphics×14.5 tok/s

What it fits on

The smallest build needs about 5.7 GB; 15 of 18 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. Linux, CPU only · 2 CPUs1 GB RAMnone
  2. NVIDIA GeForce RTX 3050 Laptop GPU4 GB on the cardnone
  3. NVIDIA GeForce GTX 16504 GB on the cardnone
  4. AMD Radeon(TM) Vega 8 Mobile Graphics7 GB RAMevery build
  5. Windows, CPU only · 12 CPUs7 GB RAMevery build
  6. Apple silicon8 GB RAMevery build
  7. Apple M18 GB RAMevery build
  8. Mac, CPU only · 8 CPUs8 GB RAMevery build
  9. NVIDIA GeForce RTX 3080 Ti12 GB on the cardevery build
  10. Tesla T415 GB on the cardevery build
  11. Intel(R) UHD Graphics 73016 GB RAMevery build
  12. Intel(R) Iris(R) Xe Graphics16 GB RAMevery build
  13. Apple M416 GB RAMevery build
  14. Apple M116 GB RAMevery build
  15. NVIDIA L422 GB on the cardevery build
  16. 2× Tesla T429 GB on the cardevery build
  17. Linux, CPU only · 4 CPUs31 GB RAMevery build
  18. NVIDIA RTX PRO 6000 Blackwell Server Edition95 GB on the cardevery 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, 3 because the machine could not hold the prompt.

machine · build4K8K16K32K64K128Kwhy it stopped
Gram ·dae679NVIDIA L4 · Q4_K_M by bartowski8K recalled at every size asked
Gram ·12954dNVIDIA GeForce GTX 1650 · Q4_K_M by bartowski8K an older probe: where it stopped was not recorded
Gram ·84bf54NVIDIA GeForce RTX 3080 Ti · Q4_K_M by bartowski4K this machine could not serve at 8K
Gram ·a5be462× Tesla T4 · Q4_K_M by bartowski4K the model lost the needle at 8K
Gram ·9e3c79Intel(R) UHD Graphics 730 · Q4_K_M by bartowski4K this machine could not serve at 8K
Gram ·2e5c02Intel(R) Iris(R) Xe Graphics · Q4_K_M by bartowski4K too slow to answer here
Gram ·afbb8aApple M4 · Q4_K_M by bartowski— an older probe: where it stopped was not recorded
Gram ·a5be462× Tesla T4 · BF16 by sarvamai— the model lost the needle at 4K

Carbon per million tokens

Lowest build on each machine · 3 of 4 read off a power sensor, the rest modelled from the parts · fewer grams is better

  1. Gram ·dae679196 g1.49 J/tok · sensor
  2. Gram ·a5be46204 g1.91 J/tok · sensor
  3. Gram ·84bf54876 g4.45 J/tok · sensor
  4. Gram ·9e3c791.5 kg7.58 J/tok · modelled

Carbon is location-based: signed energy × the yearly average of the grid where the machine signed its place (world grid average (Singapore not in the table) · United States grid average · India grid average; 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 · bartowski6.0 GB8K178.28
BF16 by sarvamaisarvamai/sarvam-1vllmvendor · sarvamai5.7 GB——1