MorphMorph
Morph Ultra

Unlimited Kimi K3.1 and DeepSeek V4.1 Flash for $200 a month

One flat price for coding agents. No token meter on either model. Each subscription has per-minute rate limits on uncached input and output tokens for each model, and requests can be rate limited when we are at capacity.
Get UltraRead the docs
Morph Ultra
$200/month
Kimi K3.1
Unlimited
DeepSeek V4.1 Flash
Unlimited
DeepSeek V4.1 Flash Abliterated
Unlimited, if approved
Other models
Per token
Available to
Individuals and teams
Subscriptions
1,000 available
Rate limits, tokens per minute
ModelUncached inputOutput
Kimi K3.124,2906,162
DeepSeek V4.1 Flash400,00070,000
Plus a burst reserve of 3x each limit, refilled over 30 minutes.
Median speed, Nov 9, 23:00 UTCHourly charts
ModelTokens/secFirst token
Kimi K3.11291.67 s
DeepSeek V4.1 Flash1560.74 s
Morph Ultra

One price, no token meter

What the plan covers

Kimi K3.1 and DeepSeek V4.1 Flash, with no token meter and no monthly token cap. Every other Morph model bills per token at list price, from your credits first, then your card.

Docs
01
// Included. No token meter.
02
morph-kimik31                 // Kimi K3.1
03
morph-dsv41flash              // DeepSeek V4.1 Flash
04
morph-dsv41flash-1017         // October 17 snapshot
05
morph-dsv41flash-abliterated  // with privileged access
06
07
// Without the plan, per 1M tokens
08
// Kimi K3.1           $2.50 in, $0.29 cached, $14.00 out
09
// V4.1 Flash          $0.12 in, $0.008 cached, $0.80 out

Rate limits

Ultra removes the token meter. It does not reserve capacity. There is no limit on total tokens, only on how fast you use them. Each subscription has per-minute rate limits on uncached input and output tokens for each model, so no account can crowd out the rest. Cached input does not count toward them. Each limit also has a burst reserve of 3 times the limit for cold starts and subagent fan-outs; it refills over 30 minutes, so you can burst up to twice an hour. Requests can also return 429 when we are at capacity. The limits below are current figures; we can change limits and included models at any time (Terms of Service, Section 7.7).

Docs
01
// Per subscription, tokens per minute
02
// Model                Uncached input  Output
03
// Kimi K3.1            24,290          6,162
04
// DeepSeek V4.1 Flash  400,000         70,000
05
// Plus a burst reserve of 3x each limit, refilled over 30 minutes
06
07
// A 429 means retry. The SDK handles it.
08
const client = new OpenAI({
09
  apiKey: process.env.MORPH_API_KEY,
10
  baseURL: "https://api.morphllm.com/v1",
11
  maxRetries: 5,  // backs off and honors Retry-After
12
});

Point your agent at it

Morph serves the OpenAI and Anthropic wire formats. Two environment variables move Claude Code onto the plan.

Docs
01
export ANTHROPIC_BASE_URL="https://api.morphllm.com"
02
export ANTHROPIC_AUTH_TOKEN="YOUR_MORPH_API_KEY"
03
export ANTHROPIC_MODEL="morph-kimik31"
04
export ANTHROPIC_SMALL_FAST_MODEL="morph-dsv41flash"
05
claude