Abbreviations

Abbreviation Tables

Complete reference for M2M Protocol Token compression mappings.

Key Abbreviations

Request Keys

Original	Abbreviated	Context
messages	m	Chat messages array
content	c	Message content
role	r	Message role
model	M	Model identifier
temperature	T	Sampling temperature
max_tokens	x	Maximum output tokens
top_p	p	Nucleus sampling
stream	s	Enable streaming
stop	S	Stop sequences
n	n	Number of completions
seed	se	Random seed
user	u	User identifier
frequency_penalty	f	Frequency penalty
presence_penalty	P	Presence penalty
logit_bias	lb	Token biases
logprobs	lp	Log probabilities
top_logprobs	tlp	Top log probs count
response_format	rf	Response format
tools	ts	Tool definitions
tool_choice	tc	Tool selection
functions	fs	Function definitions
function_call	fc	Function call mode

Response Keys

Original	Abbreviated	Context
choices	C	Response choices
index	i	Choice index
message	m	Response message
delta	d	Streaming delta
finish_reason	fr	Completion reason
usage	U	Token usage
prompt_tokens	pt	Input tokens
completion_tokens	ct	Output tokens
total_tokens	tt	Total tokens
logprobs	lp	Log probabilities
created	cr	Timestamp
object	o	Object type
system_fingerprint	sf	System fingerprint

Tool/Function Keys

Original	Abbreviated	Context
tool_calls	tc	Tool call array
function	fn	Function definition
name	n	Function name
arguments	a	Function arguments
type	t	Tool type
description	desc	Tool description
parameters	params	Function parameters
required	req	Required parameters
properties	props	Parameter properties

Value Abbreviations

Role Values

Original	Abbreviated
system	s
user	u
assistant	a
function	f
tool	t

Finish Reason Values

Original	Abbreviated
stop	s
length	l
tool_calls	tc
content_filter	cf
function_call	fc

Response Format Types

Original	Abbreviated
text	t
json_object	j
json_schema	js

Model Abbreviations

OpenAI Models

Original	Abbreviated
gpt-4o	4o
gpt-4o-mini	4om
gpt-4o-2024-11-20	4o1120
gpt-4o-2024-08-06	4o0806
gpt-4-turbo	4t
gpt-4-turbo-preview	4tp
gpt-4	4
gpt-4-32k	432k
gpt-3.5-turbo	35t
gpt-3.5-turbo-16k	35t16k
o1	o1
o1-mini	o1m
o1-preview	o1p
o3	o3
o3-mini	o3m

Meta Llama Models

Original	Abbreviated
meta-llama/llama-3.3-70b	ml3370
meta-llama/llama-3.3-70b-instruct	ml3370i
meta-llama/llama-3.1-405b	ml31405
meta-llama/llama-3.1-405b-instruct	ml31405i
meta-llama/llama-3.1-70b	ml3170
meta-llama/llama-3.1-70b-instruct	ml3170i
meta-llama/llama-3.1-8b	ml318
meta-llama/llama-3.1-8b-instruct	ml318i

Mistral Models

Original	Abbreviated
mistralai/mistral-large	mim-l
mistralai/mistral-large-latest	mim-ll
mistralai/mistral-medium	mim-m
mistralai/mistral-small	mim-s
mistralai/mixtral-8x7b	mimx87
mistralai/mixtral-8x22b	mimx822
mistralai/codestral-latest	micodl

DeepSeek Models

Original	Abbreviated
deepseek/deepseek-v3	ddv3
deepseek/deepseek-r1	ddr1
deepseek/deepseek-coder	ddc
deepseek/deepseek-chat	ddchat

Qwen Models

Original	Abbreviated
qwen/qwen-2.5-72b	qq2572
qwen/qwen-2.5-32b	qq2532
qwen/qwen-2.5-coder-32b	qqc32

Default Values

Parameters with these values MAY be omitted during compression.

Parameter	Default Value
temperature	1.0
top_p	1.0
n	1
stream	false
frequency_penalty	0
presence_penalty	0
logit_bias	{}
stop	null
logprobs	false

Compression Example

Original Request

{
  "model": "gpt-4o",
  "messages": [
    {"role": "system", "content": "You are helpful."},
    {"role": "user", "content": "Hello!"}
  ],
  "temperature": 1.0,
  "max_tokens": 100,
  "stream": false
}

Compressed Request

#T1|{"M":"4o","m":[{"r":"s","c":"You are helpful."},{"r":"u","c":"Hello!"}],"x":100}

Transformations Applied

1. model → M
2. gpt-4o → 4o
3. messages → m
4. role → r
5. system → s
6. content → c
7. user → u
8. max_tokens → x
9. temperature: 1.0 → omitted (default)
10. stream: false → omitted (default)