OpenChat 3.5

OpenChat · openchat/openchat-3-5-0106

ProviderOpenChat
Familyopenchat
Typellm-chat
Statusactive
Released2024-01-07
Updated2024-05-18
Parameters7.2B
Open weightsyes

Lineage

RelationshipModel
Is a finetune ofmistralai/Mistral-7B-v0.1

What it fits on

Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.

Every figure here is computed, not measured. Fit is weights at each quantisation against device memory, with a 25% allowance for the KV cache, activations and the OS. Decode rate is the memory-bandwidth roofline at 70% efficiency. Nobody has run this model on these devices.
DeviceMemoryBandwidthBest qualityWeightstok/sFastest
Cerebras WSE-344.0 GB21000000.0 GB/sbf1614.48 GB~1014948.3~4059793.0 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf1614.48 GB~773.3~3093.2 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf1614.48 GB~386.6~1546.6 q4
NVIDIA B200180.0 GB7700.0 GB/sbf1614.48 GB~372.1~1488.6 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf1614.48 GB~356.7~1426.7 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf1614.48 GB~290.0~1159.9 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf1614.48 GB~256.2~1024.6 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf1614.48 GB~232.0~928.0 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf1614.48 GB~188.5~754.0 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf1614.48 GB~161.9~647.6 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf1614.48 GB~154.7~618.6 q4
Google TPU v5p95.0 GB2765.0 GB/sbf1614.48 GB~133.6~534.5 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf1614.48 GB~98.5~394.2 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf1614.48 GB~96.7~386.6 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf1614.48 GB~86.6~346.4 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf1614.48 GB~79.2~316.7 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf1614.48 GB~77.3~309.3 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf1614.48 GB~75.2~300.6 q4
Google TPU v432.0 GB1200.0 GB/sbf1614.48 GB~58.0~232.0 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf1614.48 GB~48.7~194.9 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf1614.48 GB~46.4~185.6 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sfp87.24 GB~92.8~185.6 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf1614.48 GB~45.2~181.0 q4
NVIDIA L40S48.0 GB864.0 GB/sbf1614.48 GB~41.8~167.0 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf1614.48 GB~39.6~158.3 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sbf1614.48 GB~38.7~154.7 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf1614.48 GB~38.7~154.7 q4
Google TPU v5e16.0 GB800.0 GB/sfp87.24 GB~77.3~154.7 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sfp87.24 GB~71.1~142.3 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sfp87.24 GB~65.0~129.9 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sfp87.24 GB~61.9~123.7 q4
Apple M4 Max128.0 GB546.0 GB/sbf1614.48 GB~26.4~105.6 q4
Apple M1 Max64.0 GB400.0 GB/sbf1614.48 GB~19.3~77.3 q4
Apple M2 Max96.0 GB400.0 GB/sbf1614.48 GB~19.3~77.3 q4
Apple M3 Max128.0 GB400.0 GB/sbf1614.48 GB~19.3~77.3 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sfp87.24 GB~34.8~69.6 q4
NVIDIA L424.0 GB300.0 GB/sbf1614.48 GB~14.5~58.0 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sfp87.24 GB~27.8~55.7 q4
Apple M4 Pro64.0 GB273.0 GB/sbf1614.48 GB~13.2~52.8 q4
Apple M432.0 GB120.0 GB/sbf1614.48 GB~5.8~23.2 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging Faceopenchat/openchat-3.5-0106

Verified benchmark evidence

Each row was checked against its source by a reviewer, and carries the model identifier as actually evaluated — which is not always the same as this card's.
BenchmarkEvaluated asScoreEvidence dateSource kind
gpqa_diamondOpenChat 3.523.03%2026-09-10 evaluatedindependent evaluatorsource

Reported benchmark scores

This card reports no benchmark scores yet.

BenchmarkCatalogue standingScore

Data

This card as JSON · See it in the graph · Edit on GitHub