Home → Help
tool_choice to auto and say in the prompt when the tool applies. On models where reasoning is always on, a forced tool call would have to skip that reasoning, so the API refuses the request rather than quietly degrading the model.400 on requests with tool_choice: {"type":"any"}400 on tool_choice: {"type":"tool","name":"..."}The same request works with tool_choice omittedAn agent framework that worked on one model fails on every call after a swapF
o
r
c
i
n
g
a
t
o
o
l
c
a
l
l
i
s
a
r
e
a
s
o
n
a
b
l
e
t
h
i
n
g
t
o
w
a
n
t
.
I
t
i
s
h
o
w
m
a
n
y
f
r
a
m
e
w
o
r
k
s
g
u
a
r
a
n
t
e
e
s
t
r
u
c
t
u
r
e
d
o
u
t
p
u
t
:
d
e
f
i
n
e
o
n
e
t
o
o
l
,
f
o
r
c
e
i
t
,
p
a
r
s
e
t
h
e
a
r
g
u
m
e
n
t
s
.
T
h
a
t
t
r
i
c
k
s
t
o
p
s
w
o
r
k
i
n
g
o
n
m
o
d
e
l
s
t
h
a
t
r
e
a
s
o
n
b
e
f
o
r
e
e
v
e
r
y
r
e
s
p
o
n
s
e
,
b
e
c
a
u
s
e
t
h
e
f
o
r
c
e
d
p
a
t
h
b
y
p
a
s
s
e
s
t
h
e
s
t
e
p
y
o
u
a
r
e
p
a
y
i
n
g
f
o
r
.
T
h
e
r
e
f
u
s
a
l
i
s
b
e
t
t
e
r
t
h
a
n
t
h
e
a
l
t
e
r
n
a
t
i
v
e
.
A
n
A
P
I
t
h
a
t
a
c
c
e
p
t
e
d
t
h
e
p
a
r
a
m
e
t
e
r
a
n
d
i
g
n
o
r
e
d
i
t
w
o
u
l
d
l
e
a
v
e
y
o
u
b
e
l
i
e
v
i
n
g
y
o
u
h
a
d
c
o
n
s
t
r
a
i
n
e
d
t
h
e
m
o
d
e
l
w
h
e
n
y
o
u
h
a
d
n
o
t
—
a
b
u
g
t
h
a
t
s
u
r
f
a
c
e
s
m
u
c
h
l
a
t
e
r
,
i
n
p
r
o
d
u
c
t
i
o
n
,
a
s
o
c
c
a
s
i
o
n
a
l
u
n
p
a
r
s
e
a
b
l
e
o
u
t
p
u
t
.
T
h
e
s
a
m
e
m
o
d
e
l
s
u
s
u
a
l
l
y
r
e
j
e
c
t
n
o
n
-
d
e
f
a
u
l
t
<
c
o
d
e
>
t
e
m
p
e
r
a
t
u
r
e
<
/
c
o
d
e
>
,
<
c
o
d
e
>
t
o
p
_
p
<
/
c
o
d
e
>
a
n
d
<
c
o
d
e
>
t
o
p
_
k
<
/
c
o
d
e
>
,
a
n
d
d
o
n
o
t
s
u
p
p
o
r
t
p
r
e
f
i
l
l
i
n
g
t
h
e
a
s
s
i
s
t
a
n
t
t
u
r
n
.
C
o
d
e
p
o
r
t
e
d
f
r
o
m
a
n
o
l
d
e
r
m
o
d
e
l
i
n
t
h
e
s
a
m
e
f
a
m
i
l
y
h
i
t
s
s
e
v
e
r
a
l
o
f
t
h
e
s
e
a
t
o
n
c
e
,
w
h
i
c
h
m
a
k
e
s
i
t
l
o
o
k
l
i
k
e
a
n
a
u
t
h
e
n
t
i
c
a
t
i
o
n
o
r
e
n
d
p
o
i
n
t
p
r
o
b
l
e
m
r
a
t
h
e
r
t
h
a
n
a
p
a
r
a
m
e
t
e
r
o
n
e
.
Send the same request twice, changing only tool_choice:
for CHOICE in '{"type":"auto"}' '{"type":"any"}'; do
echo -n "$CHOICE -> "
curl -s -o /dev/null -w '%{http_code}\n' \
'YOUR_BASE_URL/chat/completions' \
-H 'Authorization: Bearer YOUR_KEY' \
-H 'Content-Type: application/json' \
-d "{\"model\":\"YOUR_MODEL\",\"max_tokens\":16,
\"messages\":[{\"role\":\"user\",\"content\":\"hi\"}],
\"tool_choice\":$CHOICE}"
doneIf auto returns 200 and any returns 400, the model does not support forced tool use. Nothing about your key or endpoint is wrong.
claude-fable-5-1 and claude-fable-5: forced tool use, non-default sampling parameters and assistant prefill are all rejected. The Fable 5.1 guide has the full list alongside its cache pricing.Last checked 2026-09-09. Written from problems diagnosed on a live OpenAI-compatible gateway, not collected from other sites.